fleetd #778: never nudge a pane to run a call it may not perform

ReplyPushLoop nudged an observer pane to run fleet_poll(ticket=...) five
times, which Authz refused every time -- the pane was the right one to
nudge, but the instruction was one it could never follow. Gate ticket
nudges generally: ReplyPushLoop takes a (lead, ticket) -> boolean
authorization predicate, consulted at the one place every other lookup in
the class reads from (pendingTicketsFor), so a ticket the resolved lead may
not poll is invisible to decide/injectNudge/bumpNudgeCounts alike, never
named in a nudge, and never spends its own nudge budget.

The real predicate, wired in FleetdAssembly, reconstructs the resolved
lead's Principal from its bare terminal (CallerResolver.resolveTerminal,
extracted from the existing connection-based resolve()) and runs it through
the same Authz.permits/MessageService.isTicketOwnedBy check the real
fleet_poll call would face. Reply and question nudges need no equivalent
gate: Authz already restricts who can delegate to a worker in the first
place (SPAWN and SEND-to-a-worker are primary/architect-only), so the lead
resolved for those two sources is always one DRAIN/ANSWER already grants
unconditionally.

Also fixes the nudge-sent/failed log lines, which called any ticket creator
a "lead" regardless of its actual role.

Test: ReplyPushLoopTest proves a forbidden ticket produces no nudge, paired
with a positive control proving the same setup nudges normally once the
predicate authorizes the call.

.claude/skills/handover/SKILL.md

@@ -13,7 +13,7 @@ writes a handover file, and the new session reads that file and carries on.
 - **By hand.** You write the file, then tell the operator where it is. The operator starts the new
   session and points it at the file. This always works.
 - **With `fleet_handover`** (fleetd #480, merged 2026-09-11). You ask fleetd to do the swap: it
-  checks the file, clears your pane, and tells the fresh session to read it. This needs
+  checks the file, restarts your pane, and tells the fresh session to read it. This needs
   `leadRollover:` in `fleetd.yaml`; without it every action answers a clean refusal naming
   `NOT_CONFIGURED`, and you fall back to the manual path. Section 11 below is the procedure.
 
@@ -133,8 +133,17 @@ A handover file is a record of state and decisions. It is not a diary.
 **Run the three steps in this order. The order is not a style choice — the wrong order is
 refused.**
 
-1. **`fleet_handover{action: "open", reason: "<why now>"}`.** It returns a `token` and the
-   `handoverPath` you must write to. Nothing has happened to your pane yet.
+1. **`fleet_handover{action: "open", reason: "<why now>"}`.** It returns a `token`, the
+   `handoverPath` you must write to, and `outstandingTickets` plus `openAsks`. Nothing has happened
+   to your pane yet.
+
+   **Copy `outstandingTickets` and `openAsks` into the handover file.** Your successor keeps the
+   authority to poll those tickets and answer those asks, because both are gated on the lead's
+   name, which does not change when your pane does. What it does not keep is the ids — they exist
+   only in your context and in this response. A ticket already in a terminal phase is the urgent
+   one: its reply lives only in memory and is deleted once the ticket TTL passes, so an uncollected
+   report is lost for good. Poll those before you confirm, or name them in the file so your
+   successor polls them first.
 
    **Write to exactly that path, and do not resolve it yourself.** It is always absolute, even when
    the operator configured a relative `handoverPath`: fleetd resolves a relative one against your
@@ -159,15 +168,57 @@ fails.
 
 `{action: "cancel", token}` drops a pending request without rolling.
 
-**Things that will surprise you:**
+## 12. A roll restarts your process
 
-- **`accepted` does not mean your pane has been cleared.** It means every gate passed and the roll
+The daemon ends your pane, launches a fresh one, waits for the new terminal to be recognised as a
+lead, and only then sends the bootstrap text. Your `claude` process really exits, so a newer CLI on
+disk is loaded. It does not type `/clear`.
+
+**The restart path is live in the code but has not run yet.** Measured 2026-10-05: the log holds no
+`lead-rollover:` line since the current daemon started, and no occurrence of any new outcome name.
+All 20 rolls recorded further down ran under the older `/clear` behaviour, so read them as history
+rather than as evidence about your own roll. Re-measure with:
+
+```bash
+grep -c "lead-rollover: rolled" fleetd/fleetd.out   # successful rolls
+grep -c "lead-rollover:" fleetd/fleetd.out          # positive control: must be larger
+```
+
+Run the control line too. A broken pattern returns a clean `0` that reads exactly like good news.
+If the first number has grown past 20, somebody has rolled under the restart path, and this section
+should be replaced with what they measured.
+
+**Three separate timeouts bound a roll.** `leadRollover.relaunchReadySeconds` (default 45,
+`FleetConfig.java:1483`) bounds **each** of two waits that run after the relaunch, so the worst case
+there is about twice that number, not 45 seconds in total. A third bound gives your old pane 10
+seconds to die (`LeadRollover.PANE_DEATH_TIMEOUT_SECONDS`).
+
+`fleet_handover{action: "status", token}` answers with one of these:
+
+| Outcome | What it means |
+|---|---|
+| `IN_PROGRESS` | still running; it always ends on one of the rows below |
+| `ROLLED` | the roll succeeded |
+| `TURN_NEVER_SETTLED` | your turn ran past `leadRollover.turnSettleSeconds`; nothing was touched |
+| `OLD_PANE_NEVER_DIED` | your pane did not exit within the 10-second bound |
+| `RELAUNCH_FAILED` | launching the fresh pane failed |
+| `RELAUNCH_NEVER_READY` | the fresh pane never became ready within `relaunchReadySeconds` |
+| `RELAUNCH_NOT_RECOGNISED` | the fresh terminal never resolved as a lead |
+| `FAILED` | the roll threw; `runRollover`'s catch records this rather than leaving it stuck |
+
+Only `TURN_NEVER_SETTLED` guarantees your context is intact. The other failures can leave you
+already gone, so you may never read them yourself — they are in the daemon log for whoever looks
+next.
+
+## 13. Things that will surprise you
+
+- **`accepted` does not mean your pane has been restarted.** It means every gate passed and the roll
   is scheduled to run once your current turn ends. Say your goodbye in the same turn — you will not
   get another one.
-- **If you are still running after that turn, the roll did not happen.** A roll that works clears
-  you, so surviving your own goodbye is itself the signal that it refused. Check with
+- **If you are still running after that turn, the roll did not happen.** A roll that works ends your
+  process, so surviving your own goodbye is itself the signal that it refused. Check with
   `fleet_handover{action: "status", token}`, using the token you confirmed. `TURN_NEVER_SETTLED`
-  means your turn ran past `leadRollover.turnSettleSeconds` and **no `/clear` was ever sent**: your
+  means your turn ran past `leadRollover.turnSettleSeconds` and **your pane was never ended**: your
   context is intact and nothing was lost. Open a fresh request and retry. Never assume the roll
   succeeded because `confirm` answered `accepted` — by the time it refuses, there is no caller left
   to tell, so this check is the only thing that closes that gap.
@@ -193,41 +244,29 @@ fails.
   **deferred, not hot** — it is read once at boot, so an edit does nothing until the daemon is
   redeployed.
 
-  **Until fleetd #621 merges, the nudge text will tell you to ask the operator even where the
-  daemon no longer requires it.** `LeadHeartbeatLoop.contextNotice()` hardcodes "ask the operator"
-  and takes no config, so it cannot know. Trust the config value over the nudge text. Once #621 is
-  merged and deployed, the nudge matches the config and this warning can be deleted.
+  The context nudge tracks this value, so its text and the config agree (fleetd #621 —
+  `LeadHeartbeatLoop.contextNotice` takes `requireOperatorConfirm`). You still read the config
+  rather than the nudge, because the nudge only reaches you when your context is already high.
 
 - **The roll can still refuse after `confirm` returns**, and by then there is no caller to tell.
   Those outcomes are logged only, as `lead-rollover:` lines in the daemon log.
-- **The bootstrap prompt works end to end. Measured 2026-09-22.** This used to say the fix was
-  unproven (fleetd #489) and told you to expect a failure. That is no longer true. The daemon log
-  now holds four `lead-rollover: rolled` lines, and three of them ran on 2026-09-22 at 10:01:43,
-  10:38:28 and 11:15:47. Each one cleared the old lead and started a fresh session against the
-  handover file, with the configured `bootstrapText` arriving as its first message. No context was
-  lost. The old `Unknown command: /clearFresh` failure from 2026-09-12 does not appear in the log
-  at all. Re-measure both numbers with:
+- **The bootstrap prompt works end to end — measured under the older `/clear` path.** On 2026-09-22
+  the daemon log held four `lead-rollover: rolled` lines; on 2026-10-04 it held **20**, against a
+  control of 86 `lead-rollover:` lines. Each roll started a fresh session against the handover file,
+  with the configured `bootstrapText` arriving as its first message, and no context was lost. So the
+  bootstrap half of the roll is proven, and that half did not change. Section 12 says how to check
+  whether anything has rolled under the restart path since.
 
-  ```bash
-  grep -c "lead-rollover: rolled" fleetd/fleetd.out   # successful rolls
-  grep -c "lead-rollover:" fleetd/fleetd.out          # positive control: must be larger
-  grep -c "Unknown command" fleetd/fleetd.out         # the old failure: expect 0
-  ```
-
-  Run the control line too. A broken pattern returns a clean `0` that reads exactly like good news.
-  If the first number stops growing across rolls, or `Unknown command` returns anything above 0,
-  the bootstrap has regressed and this paragraph is stale again.
+  19 of the 20 carry an `elapsedMs`: median 16507 ms, maximum 48261 ms, and two above 45000 ms. That
+  figure times the **whole** roll, and the wait for your own turn to end dominates it. Expect a roll
+  to take tens of seconds, and do not treat a slow one as a failed one. The restart path adds a pane
+  death and a relaunch to that work, so expect it to be slower rather than faster — but nobody has
+  measured it, so do not quote a number for it.
 
   **You still write the file before you confirm, and never the other way round.** That order is not
   about the bootstrap being unreliable. It is what the daemon checks: the handover file must have
   been modified *after* the open request, or `confirm` refuses it as stale.
 
-- **One warning in the log is normal and is not a failure.** Every one of the three rolls above also
-  logged `/clear on term_… was never observed as WORKING after 8 consecutive IDLE/DONE polls —
-  releasing rather than wedging the roll`. The daemon could not see the pane go WORKING after
-  `/clear`, so it released instead of hanging. The roll then succeeded anyway. That is the safe
-  branch behaving correctly. Do not report it as a broken roll.
-
 ## Writing style
 
 Write in plain English. Use everyday words, one idea per sentence, and active voice. Keep every

.claude/skills/redeploy-fleetd/SKILL.md

@@ -59,9 +59,12 @@ if the script is unavailable or a step fails, this is what it was protecting you
 3. **A restart is the only way deferred config keys take effect.** That is usually the reason to do
    it. The startup log names which keys it accepted and which it deferred — read those lines rather
    than assuming.
-4. **Re-check identity afterwards.** Call `fleet_whoami` and confirm it still answers `primary`. The
-   lead is found by its tab label (`fleet.leaders.*.tab`), and a lead whose tab no longer matches is
-   demoted to worker, which refuses every orchestration call.
+4. **Re-check identity afterwards.** Call `fleet_whoami` and confirm it still answers `primary`. A
+   lead is found by two things together: its tab is labelled `lead`, and that tab sits in the space
+   named by `fleet.leaders.<name>.workspace`. Both must match, so a renamed tab *and* a space whose
+   label differs from the config each demote the lead to worker, which refuses every orchestration
+   call. A `tab:` still in config is accepted as a second label for that lead, and the daemon logs
+   one deprecation warning naming it at startup.
 5. **Prove the new jar is the one running.** Confirm a *fresh* `fleetd listening` line at the end of
    `fleetd/fleetd.out`, dated after the restart. An old daemon that never died looks identical from
    the outside.

CLAUDE.md

@@ -27,23 +27,37 @@ through its `fleet_*` tools. No session addresses a peer, a broker, or the netwo
 **Every role reads this file.** A member runs in a git worktree of this same repo, so it inherits
 this `CLAUDE.md` verbatim, and every rule below is role-conditional.
 
-**Call `fleet_whoami`.** It returns `primary`, `worker`, or `architect`, resolved by the daemon from
-your connection — unforgeable, and the same resolution its authorization gate uses. A worker also
-carries its `sessionId`, `profile`, `worktree` and `branch`; an architect carries the slot name it
-was bound to. Don't infer what you can ask.
+**Call `fleet_whoami`.** It returns `primary`, `worker`, `architect`, `collaborator`, or `observer`,
+resolved by the daemon from your connection — unforgeable, and the same resolution its authorization
+gate uses. A worker also carries its `sessionId`, `profile`, `worktree` and `branch`; an architect
+carries the slot name it was bound to; a collaborator carries its registry name and its own
+`sessionId`, and **no `leader` key** — a collaborator is a named peer, not a primary. An **observer**
+carries only its own `sessionId`: a pane the daemon could not place as any of the above, authorized
+to `READ`/`METRICS`, to `REPLY`/`ASK` on its own pane, and to `SEND` only to a target that resolves
+as an observer too — never to a lead, a collaborator, or a spawned member, and never a ticket. It
+finds such a target in `fleet_list`'s `panes` array, which for an observer is filtered to exactly
+what it may send to and reduced to `sessionId`, `label`, `status`, `role` and `deliverable`.
+Don't infer what you can ask.
 
 Only if that call is unavailable, fall back to these — each is one-way, so keep reading until one
 fires: the reply charter in your system prompt (*"You are a spawned member in the
 claude-bridge fleet"*) ⇒ **spawned member**; fleet tools prefixed `mcp__fleet__*` ⇒ **spawned
 member** (the launcher fixes that mount name; a primary's mount is named by whoever wrote its
 `.mcp.json`, so it varies — and a member spawned before CB-632 still says `mcp__bridge__*`); `ANTHROPIC_BASE_URL` set ⇒ **spawned member** (Claude-model members run
-on a clean env, so its *absence* proves nothing). None of these separate a worker from an architect —
-only `fleet_whoami` does. **Still unsure ⇒ act as a worker**, the most restricted member role. The
-two mistakes are not symmetric: a primary acting as a worker is refused by the authorization gate —
-loud and self-correcting — while a member acting as the primary ends its turn with no `fleet_reply`,
+on a clean env, so its *absence* proves nothing). None of these separate a worker from an architect,
+or a worker from an **observer** — an observer is just as unspawned as a collaborator and carries
+none of these signals either, so only `fleet_whoami` tells the two apart. **And none of them fires
+for a collaborator at all**: every signal in the ladder detects a *spawned* member, while a
+collaborator is a tab a person opened by hand, so it has no charter, no fixed mount name and a
+normal environment. A collaborator — or an observer — that cannot call `fleet_whoami` therefore falls
+to the line below and acts as a worker. That is the safe direction — it under-privileges, and the
+refusals are loud — but it means a collaborator or an observer has no way to learn what it is except
+by asking. **Still unsure ⇒ act as a worker**, the most restricted member role this ladder can name.
+The two mistakes are not symmetric: a primary acting as a worker is refused by the authorization gate
+— loud and self-correcting — while a member acting as the primary ends its turn with no `fleet_reply`,
 and the sender silently receives nothing. Fail toward the recoverable error.
 
-### Invariants — both roles, no exceptions
+### Invariants — every role, no exceptions
 
 1. **Never set, export, or forward `ANTHROPIC_BASE_URL`** (or `ANTHROPIC_AUTH_TOKEN`). The primary
    stays on subscription; only the bridge puts a member off it, at spawn. Mounting the bridge must
@@ -51,13 +65,22 @@ and the sender silently receives nothing. Fail toward the recoverable error.
 2. **The bridge is the only channel.** Text you print in your terminal reaches nobody — the other
    side cannot see your screen. An answer that isn't in a `fleet_*` call is silently discarded.
 3. **Identity comes from the connection, never an argument.** Workers never pass a target; you
-   cannot act as another session. Spawn/stop/drain are lead-only; **send is lead or architect**;
-   reply/ask are only-as-itself — any peer may answer for its own pane, and for no other. A call
-   outside your role is refused, not queued.
+   cannot act as another session. Spawn/stop/drain are lead-only; **send is lead, architect,
+   collaborator, or observer** — and a collaborator may send only to a lead or another collaborator,
+   never to a spawned member's terminal, while an observer may send only to another observer pane;
+   reply/ask are only-as-itself — any peer may answer for its own pane,
+   and for no other. A call outside your role is refused, not queued.
 4. **Delivery is status-gated: one message per turn.** Don't busy-poll a peer's terminal and don't
    re-send because a call looks slow — the bridge delivers when the peer is `idle`, `blocked` or
    `done`. A spawned member must **also** have mounted the bridge MCP: until it has, it is not
    deliverable, and a send waits on that gate for ~60s and then fails without ever reaching its pane.
+   **A lead's own pane has a second gate: its input box must be empty.** The multiplexer pastes and
+   submits in one step, so a delivery that lands while the operator is typing submits their
+   half-written line. A heartbeat, a ticket nudge and lead-to-lead mail therefore wait until the box
+   is clear, and a pane the daemon cannot read as a box waits too. Nothing is lost — every one of
+   those paths retries — but a lead that leaves text sitting in its box receives nothing until it
+   clears, and the only sign is one warning in `fleetd.out` after 20 held checks in a row. Delivery
+   to a *member* is not gated this way, because nobody types in a member's pane.
 5. **Never move a fleet session, pane or peer except through the bridge.** The bridge owns policy;
    the multiplexer owns PTYs. Any route that changes fleet state without the bridge's checks
    bypasses every rule above — the `herdr` CLI and its socket are the usual example.
@@ -93,7 +116,11 @@ below are the procedure — run them in order, every task, not only the big ones
 5. **Collect** — `fleet_poll{ticket}` → `fleet_ack{target, msgId}`. Answer a worker's `fleet_ask`
    with `fleet_send{turnId, content}` — **not** `sessionId`. A worker gone quiet is diagnosed with
    `fleet_status`, never by reading its terminal; it also reports an open question and the `turnId`
-   that answers it. **A worker's ask waits ~55 seconds, and no nudge makes that longer** — so never
+   that answers it — but **only to the caller that created that delegation**, so a question raised
+   under an architect's brief is invisible to you, and seeing none does not mean there is none.
+   **Only that same creator can answer it.** A `turnId` you came by any other way is refused, so an
+   architect's worker waits for that architect and not for you.
+   **A worker's ask waits ~55 seconds, and no nudge makes that longer** — so never
    brief a worker to "ask me". Decide before you delegate, or give it an explicit default.
    **A correction cannot reach a busy member.** A `fleet_send` to a working member is *accepted* and
    returns a ticket, and is then never delivered — measured here three times in one session, and the
@@ -158,9 +185,11 @@ you decide.
 | See the fleet | `fleet_list` → `leads` (your peers) + `members` (each carries `agentSessionId` when its backend knows one) + `loopHealth` (`RUNNING`, `STALLED`, or `STOPPED` for `statusPoller` and `sessionReaper`) · one peer's state: `fleet_status{sessionId}` |
 | Delegate (blocking) | `fleet_send{sessionId, content}` |
 | Delegate (long task) | `fleet_send{sessionId, content, wait:false}` → ticket → `fleet_poll{ticket}` |
-| Answer a member's `fleet_ask` | `fleet_send{turnId, content}` — **not** `sessionId` |
+| Answer a member's `fleet_ask` | `fleet_send{turnId, content}` — **not** `sessionId`, and only the caller that created that delegation |
 | Message a **peer lead** on this host | `fleet_send{sessionId: <their terminal>, content}` — `fleet_list` → `leads` reports it. Coordination only, **never** a task |
 | Message a **peer lead** on another daemon or host | `fleet_send{coordId: <their coord-id>, content}` — needs a `coordinator:` block; your own coord-id is in `fleet_list`. Coordination only, **never** a task |
+| Message a **collaborator** on this host | `fleet_send{sessionId: <their terminal>, content}` — `fleet_list` reports a `collaborators` array, and each row carries that peer's `name` and the `sessionId` you send to. It is visible to you, to an architect and to another collaborator, never to a worker. Coordination only, **never** a task |
+| Message an **unconfigured pane** — a tab a person opened by hand | `fleet_send{sessionId: <their terminal>, content}` — it needs **no** `fleet.collaborators` entry and no restart, because a pane becomes deliverable the moment its agent connects the bridge MCP. `fleet_list`'s `panes` array reports every such pane with its label and the terminal id to send to — the full row for you, an architect or a collaborator; filtered and reduced for an observer. **`ListAgents` still never lists these**, and joining `herdr tab list` to `GET /agents` on `tab_id` stays the read-only fallback if the array is missing. Such a pane resolves as an `observer`: it can answer you with `fleet_reply`, and it can `fleet_send` to another observer pane, but never to you. Coordination only, **never** a task |
 | Answer a peer lead that messaged you | `fleet_send{coordId}` — or `{sessionId}` if they are on this host. **Not** `fleet_reply`: it has no peer route and the publish is refused |
 | Read your own held lead-to-lead mail (no ack) | `fleet_poll{coordId: <your own coord-id, from fleet_list's coordinator.selfId>}` — primary-only; never acks, so `fleet_list`'s `held[]` still shows it after. `fleet_list`'s `held[]` gives only a truncated preview — this is the only way to read the full body |
 | Collect a held reply | `fleet_poll{target}` · then `fleet_ack{target, msgId}` |
@@ -234,6 +263,27 @@ simply complies has thrown away the reason there are two of you.
 7. **Never merge.** Stage files explicitly — never `git add -A` — and leave alone anything the
    project marks as not-yours-to-commit.
 
+### Collaborator — a named peer, not a member
+
+`fleet_whoami` answered `collaborator`, so your pane's tab matches a `fleet.collaborators.<name>.tab`
+entry. You are **not** a member: nothing delegates to you, you have no brief, no worktree and no
+ticket, and **you owe no `fleet_reply`** — the turn contract above is for a session a lead spawned,
+and it does not apply to you. Read it only to understand what the members around you are doing.
+
+What you may do: observe the fleet (`fleet_list`, `fleet_profiles`, `fleet_whoami`), and send to a
+lead or to another collaborator. What you may not: spawn, stop or drain anything, roll a lead's
+session, answer a member's `fleet_ask`, poll a ticket, or send to a spawned member's terminal. Each
+of those is refused at the gate, not queued.
+
+Two limits worth knowing before you hit them. **You cannot reach a worker** — not even to help one —
+because a worker belongs to the lead that spawned it, and routing around that would make you a
+second orchestrator with no plan. Send to the lead instead. And **you cannot read a ticket**, so you
+cannot collect a delegation's reply: `fleet_poll` refuses you at the role gate, and a ticket also
+records the terminal that created it, so even a leaked id reads nothing.
+
+Being named buys you a channel, not authority. Your `fleet_send` to a lead is coordination between
+peers: the lead owes you no obedience, and you owe it none.
+
 ### Where each rule lives (don't duplicate — extend the right layer)
 
 | Layer | Scope | Reaches |
@@ -301,10 +351,22 @@ must obey belongs in the charter, not here.
   (#362). **Read `plugin/` before designing anything about onboarding a project.** Two limits are
   structural, not bugs: a plugin cannot carry the role agent files, because
   `ClaudeCodeLauncher.java:371` requires `<cwd>/.claude/agents/<role>.md` in the member's own
-  worktree; and a plugin cannot deliver anything to members at all, because
-  `ClaudeCodeLauncher.java:285` exports `CLAUDE_CONFIG_DIR` and every Claude profile here sets it,
-  so a member never reads the operator's plugin store. **The plugin is the lead-side surface;
-  member-facing assets travel in the worktree.**
+  worktree; and a plugin reaches a member only through `CLAUDE_CONFIG_DIR`, which
+  `ClaudeCodeLauncher.java:286` exports with `putIfPresent` — so only for a profile that sets
+  `configDir`. Every `claude-code` profile does set one (the four without are `opencode`, which
+  never reads that variable). **But measured 2026-10-04: two of them point at
+  `~/.ccs/instances/ltms`, which is the operator's own `CLAUDE_CONFIG_DIR` on this host.** So for an
+  `opus` or `sonnet` member, "a member never reads the operator's plugin store" is false — it reads
+  the same store, because that store is the one its `configDir` names. It stays true for `local` and
+  `local-direct`, which point at `~/.ccs/instances/gx10`. `ClaudeCodeLauncher`'s own javadoc names
+  the related hazard: that file is rewritten on every spawn, so for those two profiles fleetd and the
+  operator's live session write the same `.claude.json`, and its compare-and-swap "narrows the
+  lost-update window, it does not close it". Re-measure which profiles share the operator's dir with
+  `awk '/^profiles:/{i=1;next} /^[a-z]/{i=0} i&&/^  [a-z-]+:$/{p=$1} i&&/configDir:/{print p,$2}'
+  fleetd/fleetd.yaml` against `echo $CLAUDE_CONFIG_DIR`; delete this note once no profile names the
+  operator's dir. **Treat the plugin as the lead-side surface and put member-facing assets in the
+  worktree** — that conclusion holds either way, because a worktree asset does not depend on which
+  config dir a member reads.
 - **Never commit** `.mcp.json` (the primary's local copy, flagged `--skip-worktree`) or `wiki/`
   (a submodule with its own remote).
 - **A provisioned worktree neutralizes `.mcp.json`, `opencode.json` and `.autoenv`** — the repo's
@@ -359,7 +421,7 @@ Before you call any work done, check the row that matches what you touched:
 | `ConnectionIdentity` / how a caller is resolved | the `fleet_whoami` paragraph and the fallback ladder |
 | `REPLY_CHARTER`, or a launcher's mount/flags | the fallback ladder (`mcp__fleet__*`), and the layering table's top row |
 | the injector / status gating | invariant 4 |
-| worktree provisioning or the parity overlay | the "both roles read this file" premise — it rests on the worker's worktree being a checkout of this repo |
+| worktree provisioning or the parity overlay | the "every role reads this file" premise — it rests on the worker's worktree being a checkout of this repo |
 | `.claude/skills/**` | the addendum's skill list, and the "name the playbook" rule |
 | a new peer kind (non-Claude adapter) | what that peer can read — anything it must obey belongs in its charter, not in the block |
 | **anything an operator can use, configure, or observe** — an MCP tool, a `fleetd.yaml` knob, an endpoint, a visible behaviour | **[Features](wiki/11-Features.md)** — one entry: what it does · the knob that turns it on · **why it exists** · the gotcha |
@@ -452,3 +514,33 @@ to replace them.
 
 Prefer the unnamed lambda parameter `_` for required-but-unused params; a non-public
 `static void main(String[])` is valid (JEP 512) and boots via `java -jar`.
+
+## Code quality — five rules, and what each already cost (enforced)
+
+Measured at `7f0c4a8`: 124 main files, 36,278 lines, of which **17,850 are code** — 44% is comment,
+and only **two** files exceed 1000 *code* lines. Encapsulation and inheritance are already sound (0
+public mutable fields; 12 `extends`, 8 of them exceptions; 120 records). So there is **no Clean Code
+section, no SOLID list and no pattern catalogue** here: two architect reviews rejected those
+independently as text that would change no behaviour. These five rules are the whole standard.
+
+1. **A comment states the current contract or a current maintainer constraint — nothing else.** No
+   tickets, history, dates, measurements or review rationale; those go in the commit message or the
+   MR description. Source code only — this rule never applies to Markdown.
+2. **A javadoc block stops at 30 lines.** Longer means it is a design argument, so it moves to
+   `docs/<subject>.md` and is linked in one line. The longest here is 235 lines
+   (`config/ConfigRef.java`) and the knowledge in it is load-bearing: **move it, never delete it.**
+   This project has **no ADR** — subject pages under `docs/` are the destination.
+3. **A comment in main source never names a test class.** There are 44 such names in 76 places and
+   **2 are already dead**, because a name inside `{@code}` is invisible to the compiler and rots in
+   silence. Say what the code guarantees; the test is found by looking.
+4. **Never relieve a testing problem by reshaping production code.** `Fleetd` carries 54 static
+   factories, `MessageService` carries 7 `volatile` race hooks, and 8 tests assert on main source as
+   *text*. Make the part injectable instead. `FleetdAssembly.assembleAndStart` is 452 lines and may
+   not grow; no new source-text test may be added.
+5. **No new package cycle, and no widening of a recorded one.** Five pairs are frozen as an exact
+   edge baseline in `PackageCyclesTest` — four of them involve `msg`.
+
+Rules 1, 2, 3 and 5 have build checks, and Gitea CI runs them on every PR, so they bind members too.
+**Rule 4's judgement half has no mechanism**: a cap stops a count growing, but no test tells a good
+decomposition from a bad one. That half is a review obligation, and saying so is deliberate — a rule
+dressed as a gate it does not have is worse than an honest review item.

docs/MCP-Contract.md

@@ -182,6 +182,12 @@ Two consequences a lead feels directly:
   is fine; the message simply waits, and then restarts the member when it next goes idle.
 - **A spawned member is not deliverable until it has mounted the MCP.** Until then a send waits on
   that gate for about 60 seconds and then fails without ever reaching the pane.
+- **The same gate is what makes an unconfigured pane deliverable.** `contextExtractor` runs on
+  every MCP request, `initialize` included, and `markTrackedCallerPresent` enrols a spawned member
+  *or* an observer into `MemberPresence`; `deliverableTo` then tests presence before the lead and
+  collaborator maps. So mounting the server is the enrolment, and a tab a person opened by hand can
+  be sent to with no config and no restart. It answers with `fleet_reply` — it cannot `fleet_send`,
+  because `Authz` keeps `SEND` to a primary, an architect or a collaborator.
 
 `UNKNOWN` is deliberately neither injectable nor a pickup. A pane whose status cannot be read is
 not a pane that is safe to write to — see fleetd #176 for what happens when a gate treats an

fleetd/fleetd.example.yaml

@@ -62,11 +62,12 @@ bind:
 #
 # Leads are configured under `fleet.leaders:` — see THE FLEET further down.
 #
-# Two things stop the tab-name convention from becoming a way to claim leadership: the configured
-# member spaces are excluded from the scan, so nothing fleetd places can land in a matching tab;
-# and startup REFUSES a `tabPrefix` that the fleet tabLabel template, or any per-profile `tabLabel`
-# override, also matches — so the two namespaces cannot overlap by accident. The label is a NAME,
-# never a capability: what a pane may do is decided by the role the daemon resolves for it.
+# Two things stop the tab-name convention from becoming a way to claim leadership: startup REFUSES
+# a `tabPrefix` that the fleet tabLabel template, or any per-profile `tabLabel` override, also
+# matches, so the two namespaces cannot overlap by accident; and the CallerResolver asks the live
+# spawned-member roster BEFORE any tab map, so a live member is never mistaken for a lead no matter
+# what its tab says. The label is a NAME, never a capability: what a pane may do is decided by the
+# role the daemon resolves for it.
 
 # CB-551: IDLE-LEAD HEARTBEAT — nudge the single lead back to work when it has been continuously
 # idle (no open fleet_send driving it) past the quiet period. The fleet is one lead + architects +
@@ -98,17 +99,17 @@ bind:
 #   contextHighNudge: false
 
 # Lead rollover (fleetd #480): replace a lead session that has decided it is ready to be replaced,
-# without an operator doing it by hand. A lead writes a handover file, then asks fleetd to clear its
-# own pane and bootstrap a fresh session against that file.
+# without an operator doing it by hand. A lead writes a handover file, then asks fleetd to end its
+# own pane, launch a fresh one, and bootstrap that fresh session against the handover file.
 #
-# Opt-in on purpose — it clears the lead's own pane on request, so upgrading the daemon must never
-# acquire that ability for you. Absent block = feature off, and nothing is constructed at all. Even
-# once present, nothing but an explicit confirm() call — one that passes every check — can ever
-# cause a /clear: there is no recurring timer, heartbeat or scheduler anywhere in this feature that
-# fires one on its own initiative. confirm() itself is called FROM the calling lead's own turn, so
-# it cannot clear the pane inline (that pane is still WORKING); instead it schedules a one-shot
+# Opt-in on purpose — it tears down the lead's own pane on request, so upgrading the daemon must
+# never acquire that ability for you. Absent block = feature off, and nothing is constructed at all.
+# Even once present, nothing but an explicit confirm() call — one that passes every check — can ever
+# tear a pane down: there is no recurring timer, heartbeat or scheduler anywhere in this feature that
+# fires one on its own initiative. confirm() itself is called FROM the calling lead's own turn, so it
+# cannot act on the pane inline (that pane is still WORKING); instead it schedules a one-shot
 # continuation that waits for the SAME confirm() call's turn to end, then does the actual work. See
-# dev.ltms.fleet.lead.LeadRollover's class javadoc for the exact order (fleetd #480 correction).
+# dev.ltms.fleet.lead.LeadRollover's class javadoc for the exact order.
 #
 # handoverPath: REQUIRED when this block is present — where the handover file a fresh lead session
 #               reads must live. No default (an operator-specific path); a present block with no
@@ -117,25 +118,30 @@ bind:
 #               working directory when that lead has none configured) — never against whatever
 #               directory the daemon process happens to have been started in. An absolute path is
 #               used unchanged. Prefer an absolute path if the daemon and the lead's pane might not
-#               share a working directory (fleetd #480 follow-up).
+#               share a working directory.
 # requireOperatorConfirm: true    # default true — confirm() refuses unless the caller also passes
 #                                 # operatorConfirmed: true
 # maxDocAgeSeconds: 3600          # default 3600 — refuse a handover file older than this
 # turnSettleSeconds: 20           # default 20 — how long the deferred roll waits for the CALLING
 #                                 # lead's own turn to end (its pane to report injectable again)
-#                                 # before sending /clear at all. If this elapses, /clear is NEVER
-#                                 # sent — a lead that never goes idle is still doing real work.
-# clearSettleSeconds: 20          # default 20 — how long to wait for the pane to become injectable
-#                                 # again AFTER /clear before giving up (never sends bootstrapText
-#                                 # if this elapses). A separate, second wait from turnSettleSeconds.
+#                                 # before tearing the old pane down at all. If this elapses, nothing
+#                                 # is torn down — a lead that never goes idle is still doing real
+#                                 # work.
+# relaunchReadySeconds: 45        # default 45 — bounds two later waits, after the old pane is gone
+#                                 # and a fresh one has been launched: first, for the fresh pane to
+#                                 # reach a real turn boundary (never sends bootstrapText if THIS one
+#                                 # elapses); second, for the new terminal to be recognised as this
+#                                 # lead (bootstrapText is sent either way once the first wait
+#                                 # passes). A separate, later pair of waits from turnSettleSeconds.
 # bootstrapText: "..."            # default names the RESOLVED (absolute) handoverPath — sent to
-#                                 # the lead once its pane settles after /clear
+#                                 # the freshly relaunched lead's pane once it reaches a real turn
+#                                 # boundary
 # leadRollover:
 #   handoverPath: /path/to/handover.md
 #   requireOperatorConfirm: true
 #   maxDocAgeSeconds: 3600
 #   turnSettleSeconds: 20
-#   clearSettleSeconds: 20
+#   relaunchReadySeconds: 45
 #   bootstrapText: "Fresh lead session: read the handover file and carry on."
 
 # Fleet health detection is dormant unless enabled (CB-573). It reads one whole-fleet agent list
@@ -659,9 +665,10 @@ fleet:
   #     tabPrefix: "lead:"           # only used to guard against a worker tabLabel colliding with
   #                                  # this convention at startup; plays no part in matching a lead
   #     scanIntervalSeconds: 10      # rescan cadence, and the worst case before a new tab is seen
-  #     workspace: leads             # where a launched lead's tab is created (default "leads").
-  #                                  # MUST NOT be a member workspace — those are excluded from the
-  #                                  # scan, so a lead placed in one is never found again.
+  #     workspace: leads             # where a launched lead's tab is created (default "fleet",
+  #                                  # the same shared space the members use). Sharing that space
+  #                                  # with members is the normal shipped shape: the scanner tells
+  #                                  # a lead from a member by the exact tab label, not by workspace.
   #     cwd: /path/to/repo           # the launched lead's working directory (default: fleetd's own)
   #     kind: claude                 # descriptive; reported by fleet_whoami
   #   gpt-sol-5.6:
@@ -669,6 +676,22 @@ fleet:
   #     kind: opencode
   #     model: openai/gpt-5.6-terra
 
+  # A collaborator tab, keyed by name (fleetd #669). A pane whose tab matches resolves as the
+  # COLLABORATOR role: `fleet_whoami` answers `collaborator`, and the session may observe the fleet
+  # (`fleet_list`, `fleet_profiles`, `fleet_whoami`) and `fleet_send` to a lead or to another
+  # collaborator. It may NOT spawn, stop or drain anything, roll a lead's session, answer a
+  # member's `fleet_ask`, poll a ticket, send across hosts, or send to a spawned member's terminal.
+  # Each of those is refused at the gate rather than queued.
+  # Recognise-only, like a profile-less `leaders:` entry above: there is no `profile:`, no
+  # `instances:` and no `kind:`. `tab:` is REQUIRED and is the only field identity depends on,
+  # matched case-insensitively — the same GET-THE-VALUE-RIGHT warning above the `leaders:` block
+  # applies here too.
+  # A lead cannot discover a collaborator yet (#703): `fleet_list` has no `collaborators` key, so
+  # the collaborator must speak first, or pass on the `sessionId` its own `fleet_whoami` reports.
+  # collaborators:
+  #   reviewer-alex:
+  #     tab: "collab: alex"
+
   # architects:
   #   architect-1:
   #     profile: opus                # a strong model, on the operator's subscription

fleetd/src/main/java/dev/ltms/fleet/Fleetd.java

@@ -217,25 +217,27 @@ public final class Fleetd {
 
     /**
      * The {@link Injector}'s readiness gate (CB-534): a target is deliverable if it is a spawned
-     * member whose agent has connected the bridge MCP, <em>or</em> a lead.
+     * member whose agent has connected the bridge MCP, a lead, <em>or</em> a collaborator.
      *
      * <p>The gate exists for one reason — to hold a delivery out of a <em>spawned</em> member's boot
      * window, where herdr already reports {@code idle} but the TUI would drop an injected paste. That
      * hazard is a property of spawning. A lead is never spawned: the operator started it and named it
-     * (or labelled its tab) only once it was up, so there is no boot window to guard.
+     * (or labelled its tab) only once it was up, so there is no boot window to guard. A collaborator
+     * is the same way — a person's own tab, matched to a configured name, never spawned.
      *
-     * <p>A lead is also never enrolled in {@link MemberPresence} — {@code FleetMcp} marks presence
-     * for every spawned member (worker and architect), deliberately, since that map doubles as the
-     * member roster's availability signal and a lead counted there would show up as an available
-     * member. So without the second disjunct a lead is permanently un-deliverable: every
-     * lead→lead send sat on the gate for {@code READINESS_GRACE_POLLS} (~60s) and then failed
-     * having never been typed into the pane.
+     * <p>Neither a lead nor a collaborator is ever enrolled in {@link MemberPresence} — {@code
+     * FleetMcp} marks presence only for a worker, an architect, or the unconfigured-pane floor,
+     * never for a lead or a collaborator. So without the second and third disjuncts a lead or
+     * collaborator is permanently un-deliverable: every send to one sat on the gate for
+     * {@code READINESS_GRACE_POLLS} (~60s) and then failed having never been typed into the pane.
      *
-     * <p>The lead set is read through the supplier on each call rather than snapshotted, so a lead
-     * discovered by {@code leadScan} after startup becomes deliverable without a restart.
+     * <p>Both sets are read through their supplier on each call rather than snapshotted, so a lead or
+     * collaborator discovered by {@code leadScan} after startup becomes deliverable without a restart.
      */
-    static Predicate<String> deliverableTo(MemberPresence presence, Supplier<Map<String, String>> leads) {
-        return target -> presence.isPresent(target) || leads.get().containsKey(target);
+    public static Predicate<String> deliverableTo(MemberPresence presence, Supplier<Map<String, String>> leads,
+            Supplier<Map<String, String>> collaborators) {
+        return target -> presence.isPresent(target) || leads.get().containsKey(target)
+                || collaborators.get().containsKey(target);
     }
 
     /**
@@ -334,22 +336,6 @@ public final class Fleetd {
         }, reasonByCredential::get);
     }
 
-    /**
-     * fleetd #415 (review follow-up): package-private factory for the CB-578 stage A {@code
-     * exhaustedPattern} startup coverage line, paired explicitly with {@link
-     * CompletionResolver.UnsetMeaning#OFF} — {@code exhaustedPattern} has no fallback, so a
-     * profile with none configured really does have the classification off.
-     *
-     * <p>Extracted out of {@code main} for the same reason {@link #capacitySource} and {@link
-     * #worktreeBranchLookup} were: {@code coverage()}'s own tests ({@code CompletionResolverTest})
-     * prove it words {@code OFF} and {@link CompletionResolver.UnsetMeaning#BUILT_IN_DEFAULT}
-     * correctly when a test supplies the meaning itself — they cannot prove {@code main} pairs the
-     * right meaning with the right key, which is the actual fleetd #415 defect. <b>Measured:</b>
-     * swapping the {@code UnsetMeaning} arguments between this method and {@link
-     * #errorPatternCoverageLine} — recreating #415's defect with the two keys exchanged — compiled
-     * with 0 errors and left all 1506 existing tests green before {@code
-     * FleetdPatternCoverageLineTest} was added to catch exactly that swap.
-     */
     /**
      * fleetd #446 follow-up: the criterion-2 WARNING text — "name the fix, not just the fact" —
      * for a profile whose {@code model:} is configured. Extracted out of the {@code
@@ -384,19 +370,22 @@ public final class Fleetd {
                 + "s quarantine above is the only thing keeping new spawns off it for now";
     }
 
+    /**
+     * Package-private factory for the {@code exhaustedPattern} startup coverage line, paired
+     * explicitly with {@link CompletionResolver.UnsetMeaning#OFF} — {@code exhaustedPattern} has
+     * no fallback, so a profile with none configured really does have the classification off.
+     */
     static String exhaustedPatternCoverageLine(Set<String> allProfiles, Set<String> configuredProfiles) {
         return CompletionResolver.coverage("exhaustedPattern", CompletionResolver.UnsetMeaning.OFF,
                 allProfiles, configuredProfiles);
     }
 
     /**
-     * fleetd #415 (review follow-up): the {@code errorPattern} counterpart of {@link
-     * #exhaustedPatternCoverageLine}, paired explicitly with {@link
-     * CompletionResolver.UnsetMeaning#BUILT_IN_DEFAULT} — an unset {@code errorPattern} still runs
-     * backend-error classification against {@code CompletionResolver}'s built-in {@code
-     * BACKEND_ERROR} pattern, so the empty case is not "off". See {@link
-     * #exhaustedPatternCoverageLine}'s javadoc for the measured swap mutation this pairing guards
-     * against.
+     * The {@code errorPattern} counterpart of {@link #exhaustedPatternCoverageLine}, paired
+     * explicitly with {@link CompletionResolver.UnsetMeaning#BUILT_IN_DEFAULT} — an unset
+     * {@code errorPattern} still runs backend-error classification against
+     * {@code CompletionResolver}'s built-in {@code BACKEND_ERROR} pattern, so the empty case is
+     * not "off".
      */
     static String errorPatternCoverageLine(Set<String> allProfiles, Set<String> configuredProfiles) {
         return CompletionResolver.coverage("errorPattern", CompletionResolver.UnsetMeaning.BUILT_IN_DEFAULT,
@@ -732,8 +721,8 @@ public final class Fleetd {
      * {@code CompletionResolver} constructor call — provably untested wiring, the whole reason
      * fleetd #248 exists: dropping that one argument (passing {@code _ -> null} instead) compiled
      * clean and left every test green. Extracted here, {@code main} now calls this factory instead
-     * of building the lambda inline, and a source assertion on that call site
-     * ({@code FleetdCompletionResolverWiringTest}) proves the argument is still actually passed.
+     * of building the lambda inline, so the argument reaching the {@code CompletionResolver}
+     * constructor is a named, directly testable call rather than an inline lambda.
      *
      * <p>Takes the roster as a plain {@link Supplier} — not a {@link SessionManager} — so this is
      * directly testable with a hand-built session list; no real {@code SessionManager} (launcher,
@@ -902,6 +891,32 @@ public final class Fleetd {
         throw (T) t;
     }
 
+    /**
+     * The {@link PrimaryRegistry} lookup for "which terminal currently hosts the lead named
+     * {@code name}" — the inverse of {@code liveLeadTerminals} (terminal id → lead name), read live
+     * on every call so a lead discovered, rolled, or lost since the last call is reflected without
+     * a restart. Returns {@code null} when no currently recognised lead carries that name — a name
+     * that is not a lead at all (an architect slot, a collaborator), or a lead whose tab the scan
+     * cannot currently place (just rolled, off-host, non-herdr).
+     *
+     * @param liveLeadTerminals terminal id → lead name for every CURRENTLY recognised lead, normally
+     *                          the same {@code leads} supplier {@code main} already builds for
+     *                          {@code HerdrRouter}/{@link #leadSeatLookup}
+     */
+    static Function<String, String> currentTerminalForName(Supplier<Map<String, String>> liveLeadTerminals) {
+        return name -> {
+            if (name == null) {
+                return null;
+            }
+            for (var entry : liveLeadTerminals.get().entrySet()) {
+                if (name.equals(entry.getValue())) {
+                    return entry.getKey();
+                }
+            }
+            return null;
+        };
+    }
+
     /**
      * fleetd #480: construct the {@link LeadRollover} executor only when {@code leadRollover:} is
      * present at startup — the same presence gate {@code leadHeartbeat:} uses just above this
@@ -939,21 +954,27 @@ public final class Fleetd {
      *                          cfg.leadHeartbeat()}
      * @param leadAgents        the {@link AgentControl} instance that reaches the LEAD's pane (not
      *                          {@code memberAgents}), normally {@code router.leadAgents()}
+     * @param leadSpaces        the {@link WorkspaceControl} instance that reaches the LEAD's
+     *                          workspace, normally {@code router.leadSpaces()} — used to tear down
+     *                          a rolled lead's old pane and confirm it is gone
+     * @param launcher          starts the fresh lead a roll relaunches once the old one is gone
      * @param config            the live {@link ConfigRef}, captured only inside the returned
-     *                          supplier and the workspace lookup — never dereferenced here
+     *                          supplier and the two lookups below — never dereferenced here
      * @param liveLeadTerminals terminal id → lead NAME for every CURRENTLY recognised lead, normally
      *                          the same {@code leads} supplier {@code main} already builds for
      *                          {@code HerdrRouter}/{@link #leadSeatLookup} — never a value snapshot
      * @return a constructed {@link LeadRollover}, or {@code null} when {@code leadRollover:} is
      *         absent from the startup config
      */
-    static LeadRollover leadRollover(FleetConfig cfg, AgentControl leadAgents, ConfigRef config,
+    static LeadRollover leadRollover(FleetConfig cfg, AgentControl leadAgents,
+            WorkspaceControl leadSpaces, LeadLauncher launcher, ConfigRef config,
             Supplier<Map<String, String>> liveLeadTerminals) {
         if (cfg.leadRollover() == null) {
             return null;
         }
+        Function<String, String> leadNameForTerminal = terminal -> liveLeadTerminals.get().get(terminal);
         Function<String, String> leadWorkspace = terminal -> {
-            String leadName = liveLeadTerminals.get().get(terminal);
+            String leadName = leadNameForTerminal.apply(terminal);
             if (leadName == null) {
                 return null;
             }
@@ -961,7 +982,8 @@ public final class Fleetd {
             FleetConfig.Leader leader = fleet == null ? null : fleet.leaders().get(leadName);
             return leader == null ? null : leader.cwd();
         };
-        return new LeadRollover(leadAgents, () -> config.get().leadRollover(), leadWorkspace);
+        return new LeadRollover(leadAgents, leadSpaces, launcher, () -> config.get().leadRollover(),
+                leadWorkspace, leadNameForTerminal, liveLeadTerminals);
     }
 
     /**

fleetd/src/main/java/dev/ltms/fleet/FleetdAssembly.java

@@ -1,7 +1,9 @@
 package dev.ltms.fleet;
 
+import dev.ltms.fleet.auth.Authz;
 import dev.ltms.fleet.auth.CallerResolver;
 import dev.ltms.fleet.auth.MemberRegistry;
+import dev.ltms.fleet.auth.Principal;
 import dev.ltms.fleet.config.ConfigRef;
 import dev.ltms.fleet.config.ConfigWatcher;
 import dev.ltms.fleet.config.FleetConfig;
@@ -43,6 +45,7 @@ import dev.ltms.fleet.msg.MessageService;
 import dev.ltms.fleet.msg.Rendezvous;
 import dev.ltms.fleet.msg.ReplyInbox;
 import dev.ltms.fleet.msg.ReplyPushLoop;
+import dev.ltms.fleet.peer.MemberRole;
 import dev.ltms.fleet.peer.PeerLauncher;
 import dev.ltms.fleet.placement.BackendOutagePolicy;
 import dev.ltms.fleet.placement.BackendQuarantine;
@@ -67,6 +70,7 @@ import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.ScheduledExecutorService;
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicReference;
+import java.util.function.BiPredicate;
 import java.util.function.Function;
 import java.util.function.Predicate;
 import java.util.function.Supplier;
@@ -141,8 +145,12 @@ final class FleetdAssembly {
                 ? ports.connectHerdr(Path.of(cfg.memberHerdrSocket()))
                 : herdr;
         AtomicReference<Supplier<Map<String, String>>> leadsRef = new AtomicReference<>(Map::of);
+        // fleetd #669 Unit E: a collaborator's pane is opened by a person, exactly like a lead's,
+        // so its terminal must also route to the lead herdr daemon rather than the member one.
+        AtomicReference<Supplier<Map<String, String>>> collaboratorTerminalsRef = new AtomicReference<>(Map::of);
         HerdrRouter router = new HerdrRouter(herdr, memberHerdr,
-                target -> leadsRef.get().get().containsKey(target));
+                target -> leadsRef.get().get().containsKey(target)
+                        || collaboratorTerminalsRef.get().get().containsKey(target));
         // CB-402: one adapter per configured peer kind, fronted by a composite router. A profile's
         // `kind:` selects its adapter — claude-code (the default) and opencode partition the profile
         // set — and the composite dispatches each SPI call to the adapter that owns the profile/pane.
@@ -249,33 +257,68 @@ final class FleetdAssembly {
         if (leadTerminals.size() > 1) {
             log.info("leads: {} panes recognised {}", leadTerminals.size(), leadTerminals.values());
         }
-        // CB-531/CB-579: discover leads by the tab labels the operator writes, one scanner per
-        // configured lead's own exact `tab:` label.
+        // Discover leads by their tab labels, one scanner per configured lead's own space. fleetd
+        // #669: the same scan also recognises a configured collaborator's tab, so one herdr pass
+        // answers both.
         final Supplier<Map<String, String>> leads;
+        final Supplier<Map<String, String>> collaboratorTerminals;
         var leaders = cfg.fleet().leaders();
-        if (!leaders.isEmpty()) {
-            Map<String, String> tabToName = new LinkedHashMap<>();
+        var collaboratorsConfig = cfg.fleet().collaborators();
+        if (!leaders.isEmpty() || !collaboratorsConfig.isEmpty()) {
+            Map<String, Map<String, String>> leadLabelsBySpace = new LinkedHashMap<>();
+            Map<String, String> spaceByLeadName = new LinkedHashMap<>();
             leaders.forEach((name, leader) -> {
-                if (leader != null && leader.tab() != null && !leader.tab().isBlank()) {
-                    tabToName.put(leader.tab(), name);
+                if (leader == null) {
+                    return;
+                }
+                spaceByLeadName.put(name, leader.workspace());
+                Map<String, String> labelsHere = leadLabelsBySpace
+                        .computeIfAbsent(leader.workspace(), k -> new LinkedHashMap<>());
+                leader.acceptedLabels().forEach(label -> labelsHere.put(label, name));
+                if (leader.tab() != null && !leader.tab().isBlank()) {
+                    log.warn("lead '{}' (fleet.leaders.{}) still configures tab: \"{}\" — deprecated, "
+                                    + "the lead tab label is now fixed to '{}'",
+                            name, name, leader.tab(), FleetConfig.Leader.LEAD_TAB_LABEL);
                 }
             });
-            int scanIntervalSeconds = leaders.values().iterator().next().scanIntervalSeconds();
+            Map<String, String> collaboratorTabToName = new LinkedHashMap<>();
+            collaboratorsConfig.forEach((name, collaborator) -> {
+                if (collaborator != null && collaborator.tab() != null && !collaborator.tab().isBlank()) {
+                    collaboratorTabToName.put(collaborator.tab(), name);
+                }
+            });
+            // A collaborator-only fleet configures no `leaders:` entry to read a scan interval from
+            // — FleetConfig.Collaborator carries no scanIntervalSeconds of its own. Falling back to
+            // FleetConfig.Leader's own compact-constructor default keeps a collaborator-only
+            // deployment on the same rescan cadence as the default lead cadence, instead of
+            // inventing a second number for the same kind of scan.
+            int scanIntervalSeconds = leaders.isEmpty()
+                    ? 10
+                    : leaders.values().iterator().next().scanIntervalSeconds();
             // This must use the lead daemon: scanning member tabs would demote the lead to a worker.
-            leads = new LeadTabScanner(herdr, tabToName, Set.of(),
-                    TimeUnit.SECONDS.toNanos(scanIntervalSeconds), ports.nanoClock());
-            log.info("lead scan: tabs {} host a lead (rescan every {}s, shared fleet space)",
-                    tabToName.keySet(), scanIntervalSeconds);
+            LeadTabScanner scanner = new LeadTabScanner(herdr, leadLabelsBySpace, collaboratorTabToName,
+                    Set.of(), TimeUnit.SECONDS.toNanos(scanIntervalSeconds), ports.nanoClock());
+            leads = scanner;
+            collaboratorTerminals = scanner::collaborators;
+            log.info("lead/collaborator scan: space per lead {}, tabs {} host a collaborator "
+                            + "(rescan every {}s)",
+                    spaceByLeadName, collaboratorTabToName.keySet(), scanIntervalSeconds);
         } else {
             leads = () -> leadTerminals;
+            collaboratorTerminals = Map::of;
         }
         leadsRef.set(leads);
+        collaboratorTerminalsRef.set(collaboratorTerminals);
+
+        // Constructed unconditionally — it is cheap and side-effect free — so a LeadRollover built
+        // below can relaunch a lead even on a boot where herdr was down for the ensureLeads() call.
+        LeadLauncher leadLauncher = new LeadLauncher(router.leadAgents(), router.leadSpaces(), cfg);
 
         // CB-558: start any declared lead that is not already running. After the scanner is built,
         // and only when herdr answered — the launcher's whole safety property is that it can count
         // live leads first, and must never guess and risk a second orchestrator.
         if (herdrUp && !leaders.isEmpty()) {
-            int launched = new LeadLauncher(router.leadAgents(), router.leadSpaces(), cfg).ensureLeads();
+            int launched = leadLauncher.ensureLeads();
             if (launched > 0) {
                 log.info("lead auto-launch: {} lead(s) started", launched);
             }
@@ -341,7 +384,7 @@ final class FleetdAssembly {
         // CB-301: the manager's presence bridge records availability and drives SPAWNING → READY.
         MemberPresence presence = sessions.asPresence();
         TurnListener turnListener = Fleetd.turnListener(completion, sessions);
-        Predicate<String> deliverable = Fleetd.deliverableTo(presence, leads);
+        Predicate<String> deliverable = Fleetd.deliverableTo(presence, leads, collaboratorTerminals);
         // fleetd #556: registration is wired directly to `completion`, not folded into the
         // `turnListener` fan-out above — so it survives `sessions.onDelivered` (or any future
         // listener) throwing, regardless of call order.
@@ -363,7 +406,8 @@ final class FleetdAssembly {
                 ports.leadMailboxOpener());
         // CB-307: learn the primary's terminal from orchestration tool calls (or pin from config).
         String pinnedPrimaryTerminal = cfg.primary() != null ? cfg.primary().terminal() : null;
-        PrimaryRegistry primaryRegistry = new PrimaryRegistry(pinnedPrimaryTerminal);
+        PrimaryRegistry primaryRegistry = new PrimaryRegistry(pinnedPrimaryTerminal,
+                Fleetd.currentTerminalForName(leads));
         // CB-532: `primary.terminal` is superseded and no longer needed for either of its jobs.
         if (pinnedPrimaryTerminal != null && !pinnedPrimaryTerminal.isBlank()) {
             log.warn("primary.terminal is DEPRECATED (CB-532) and can be deleted: identity now comes "
@@ -379,8 +423,24 @@ final class FleetdAssembly {
         // CB-502: the registry is built before the service and the push loop so send/reply outcomes
         // are counted at their single funnel rather than at each of the two caller-facing surfaces.
         Metrics metrics = FleetMetrics.create(sessions, replyInbox);
+        // The authorization check below needs `messages` and `callers`, both built further down —
+        // the same construction-order cycle `pushLoopRef` above breaks, broken the same way: a
+        // mutable holder set once each exists, read lazily from inside the predicate built here.
+        AtomicReference<MessageService> messagesRef = new AtomicReference<>();
+        AtomicReference<CallerResolver> callersRef = new AtomicReference<>();
+        BiPredicate<String, String> ticketPollAuthorized = (lead, ticket) -> {
+            MessageService messages = messagesRef.get();
+            CallerResolver cr = callersRef.get();
+            if (messages == null || cr == null) {
+                return true;
+            }
+            Principal leadPrincipal = cr.resolveTerminal(lead);
+            return Authz.permits(leadPrincipal, Authz.Action.TASK_READ, ticket,
+                    Authz.NO_KNOWN_LEAD_OR_COLLABORATOR, Authz.NO_KNOWN_OBSERVER_TARGET,
+                    t -> messages.isTicketOwnedBy(t, leadPrincipal.ownerKey()));
+        };
         var pushLoop = new ReplyPushLoop(primaryRegistry, router.leadAgents(), replyInbox,
-                pushScheduler, maxReminders, backoffMs, metrics);
+                pushScheduler, maxReminders, backoffMs, metrics, ticketPollAuthorized);
         // fleetd #201 Unit 5: point the forwarding holder captured by the backendErrorSink lambda
         // above at the real push loop, now that it exists.
         pushLoopRef.set(pushLoop);
@@ -414,9 +474,11 @@ final class FleetdAssembly {
             heartbeatScheduler.shutdownNow();
         }
         // fleetd #480: lead rollover. Opt-in; absent `leadRollover:` this is never constructed.
-        LeadRollover leadRollover = Fleetd.leadRollover(cfg, router.leadAgents(), config, leads);
+        LeadRollover leadRollover = Fleetd.leadRollover(cfg, router.leadAgents(), router.leadSpaces(),
+                leadLauncher, config, leads);
         MessageService messages = new MessageService(router, injector, rendezvous, replyInbox,
                 pushLoop, metrics);
+        messagesRef.set(messages);
 
         // Health is a slow whole-fleet observer. Keep it separate from the 250ms delivery poller.
         // FOURTH of the recurring background loops to start (optional).
@@ -452,6 +514,11 @@ final class FleetdAssembly {
         ConnectionIdentity identity = new ConnectionIdentity(
                 new PaneLocator(herdr, memberHerdr), new LsofPeerPidLookup(), new LsofProcessCwdLookup());
 
+        // fleetd #669 Unit D: a live spawned member resolves as its own role, whatever a tab map
+        // says about the same terminal. fleetd #702: SessionManager.spawnedMemberRole also answers
+        // for a pane mid-teardown, not only one still in the registry — see its javadoc.
+        Function<String, MemberRole> spawnedMemberRole = sessions::spawnedMemberRole;
+
         // CB-501: one resolver behind both entry paths. Worker identity still comes from the
         // connection and is never token-gated, so enabling token mode cannot lock the fleet out.
         final CallerResolver callers;
@@ -461,13 +528,16 @@ final class FleetdAssembly {
                 throw new IllegalStateException("auth.mode=token but env var " + cfg.auth().tokenEnv()
                         + " is unset or empty — export it before starting fleetd");
             }
-            callers = CallerResolver.withLeadsAndMembers(identity, true, token, leads, members);
+            callers = CallerResolver.withLeadsAndMembers(identity, true, token, leads, members,
+                    spawnedMemberRole, collaboratorTerminals);
             log.info("auth: token mode (bearer required for non-worker callers, env {})",
                     cfg.auth().tokenEnv());
         } else {
-            callers = CallerResolver.withLeadsAndMembers(identity, false, null, leads, members);
+            callers = CallerResolver.withLeadsAndMembers(identity, false, null, leads, members,
+                    spawnedMemberRole, collaboratorTerminals);
             log.info("auth: loopback-trust (any loopback non-worker caller is the primary)");
         }
+        callersRef.set(callers);
 
         FleetMcp.QuarantineSource quarantineSource = Fleetd.quarantineSource(config, quarantine,
                 liveExhaustedPatterns, quarantineReasonByCredential);

fleetd/src/main/java/dev/ltms/fleet/auth/Authz.java

@@ -1,5 +1,7 @@
 package dev.ltms.fleet.auth;
 
+import java.util.function.Predicate;
+
 /**
  * The authorization table (CB-505), stated once and enforced on both entry paths.
  *
@@ -20,16 +22,27 @@ public final class Authz {
         SPAWN,
         /** Tear a worker peer down. */
         STOP,
-        /** Deliver a turn to a session (or answer a worker's question). */
+        /** Deliver a turn to a local session, addressed by {@code sessionId}. */
         SEND,
+        /** Resolve a worker's blocked question and resume its turn, addressed by {@code turnId}. */
+        ANSWER,
+        /** Address a peer lead on another daemon over the coordination broker, by {@code coordId}. */
+        COORD_SEND,
         /** A worker's terminal reply for its own turn. */
         REPLY,
         /** A worker's mid-turn question to the primary. */
         ASK,
         /** Collect held replies from a session's inbox. */
         DRAIN,
-        /** Read-only observation: status, roster, profiles, task polling. */
+        /** Read-only roster, profile, and identity observation: no ticket, task, or turn state. */
         READ,
+        /**
+         * Poll a ticket, or read a session's status. Open unconditionally to a primary, a worker,
+         * and an architect; open to any other caller only for a ticket it created itself (see
+         * {@link #permits(Principal, Action, String, Predicate, Predicate, Predicate)}'s
+         * {@code callerOwnsTicket} parameter).
+         */
+        TASK_READ,
         /**
          * Read (never ack) this daemon's own held lead-to-lead coordination mail (fleetd #421).
          *
@@ -54,43 +67,157 @@ public final class Authz {
     }
 
     /**
-     * Whether {@code caller} may perform {@code action} against {@code targetSession}.
+     * The fail-closed classifier: answers no for every target, so a collaborator's {@code SEND}
+     * is refused unless a caller supplies a real one. {@code CallerResolver#knownLeadOrCollaborator()}
+     * is the real one, read from the same lead and collaborator maps {@code CallerResolver#resolve}
+     * consults, so a target that classifier calls known is one {@code resolve} would actually
+     * resolve as a lead or collaborator.
+     */
+    public static final Predicate<String> NO_KNOWN_LEAD_OR_COLLABORATOR = target -> false;
+
+    /**
+     * The fail-closed classifier for an observer's {@code SEND}: answers no for every target, so
+     * the grant is refused unless a caller supplies a real one. {@code
+     * CallerResolver#sendableObserverTarget()} is the real one, read from the same maps {@code
+     * CallerResolver#resolve} consults, so a target that classifier calls known is one {@code
+     * resolve} would actually resolve as {@link Role#OBSERVER}.
+     */
+    public static final Predicate<String> NO_KNOWN_OBSERVER_TARGET = target -> false;
+
+    /**
+     * The fail-closed classifier for {@code TASK_READ}'s ticket-ownership grant: answers no for
+     * every ticket, so the grant is refused unless a caller supplies a real one backed by the
+     * ticket's recorded creator (see {@code MessageService#isTicketOwnedBy}).
+     */
+    public static final Predicate<String> NO_CALLER_OWNS_TICKET = ticket -> false;
+
+    /**
+     * Convenience form for a caller with no classifier to supply. Fails closed: a collaborator's
+     * or an observer's {@code SEND} is refused, as if no terminal were a configured lead,
+     * collaborator, or observer target, and {@code TASK_READ} for a ticket that caller did not
+     * create itself is refused too — the same decisions {@link #NO_KNOWN_LEAD_OR_COLLABORATOR},
+     * {@link #NO_KNOWN_OBSERVER_TARGET}, and {@link #NO_CALLER_OWNS_TICKET} give explicitly. Every
+     * other action's result is identical to the six-argument form's, since none of them consult
+     * any of the three classifiers.
      *
-     * @param targetSession the session id in the request path; only consulted for the worker-scoped
-     *                      actions ({@code REPLY}, {@code ASK}), ignored otherwise, may be
-     *                      {@code null}
+     * <p>Its default classifiers deny every collaborator, every observer, and every ticket a
+     * primary/worker/architect did not already have unconditionally, so a caller enforcing
+     * authorization must use the six-argument form instead.
      */
     public static boolean permits(Principal caller, Action action, String targetSession) {
+        return permits(caller, action, targetSession, NO_KNOWN_LEAD_OR_COLLABORATOR,
+                NO_KNOWN_OBSERVER_TARGET, NO_CALLER_OWNS_TICKET);
+    }
+
+    /**
+     * As {@link #permits(Principal, Action, String)}, with a real classifier for a collaborator's
+     * {@code SEND}. An observer's {@code SEND} and {@code TASK_READ}'s ticket-ownership grant
+     * still fail closed — a caller enforcing all three grants must use the six-argument form.
+     */
+    public static boolean permits(Principal caller, Action action, String targetSession,
+                                   Predicate<String> knownLeadOrCollaborator) {
+        return permits(caller, action, targetSession, knownLeadOrCollaborator,
+                NO_KNOWN_OBSERVER_TARGET, NO_CALLER_OWNS_TICKET);
+    }
+
+    /**
+     * As {@link #permits(Principal, Action, String, Predicate)}, with a real classifier for an
+     * observer's {@code SEND}. {@code TASK_READ}'s ticket-ownership grant still fails closed — a
+     * caller enforcing all three grants must use the six-argument form.
+     */
+    public static boolean permits(Principal caller, Action action, String targetSession,
+                                   Predicate<String> knownLeadOrCollaborator,
+                                   Predicate<String> knownObserverTarget) {
+        return permits(caller, action, targetSession, knownLeadOrCollaborator, knownObserverTarget,
+                NO_CALLER_OWNS_TICKET);
+    }
+
+    /**
+     * Whether {@code caller} may perform {@code action} against {@code targetSession}.
+     *
+     * @param targetSession           the session id in the request path; only consulted for the
+     *                                worker-scoped actions ({@code REPLY}, {@code ASK}), for a
+     *                                collaborator's {@code SEND}, for an observer's {@code SEND},
+     *                                and (as the ticket id) for {@code TASK_READ}, ignored
+     *                                otherwise, may be {@code null}
+     * @param knownLeadOrCollaborator whether a terminal is a configured lead or collaborator —
+     *                                consulted only for a collaborator's {@code SEND}, to confine
+     *                                it to another named peer and never a spawned member's
+     *                                terminal
+     * @param knownObserverTarget     whether a terminal is one this daemon would itself resolve as
+     *                                {@link Role#OBSERVER} — consulted only for an observer's
+     *                                {@code SEND}, to confine it to another observer pane and never
+     *                                a lead, a collaborator, or a spawned member
+     * @param callerOwnsTicket        whether {@code targetSession} (read as a ticket id) was
+     *                                created by the calling session — consulted only for
+     *                                {@code TASK_READ} by a caller that is none of primary, worker,
+     *                                or architect, so a collaborator or an observer may read a
+     *                                ticket it created itself and no other
+     */
+    public static boolean permits(Principal caller, Action action, String targetSession,
+                                   Predicate<String> knownLeadOrCollaborator,
+                                   Predicate<String> knownObserverTarget,
+                                   Predicate<String> callerOwnsTicket) {
         if (caller == null || caller.isAnonymous()) {
             return false; // authenticated as nothing ⇒ authorized for nothing
         }
         return switch (action) {
-            // Fleet lifecycle is the primary's alone — spawn, stop, drain. An architect
-            // deliberately does NOT get these (CB-548), so it cannot tear down or stand up workers
-            // even though it coordinates them; and a worker driving any of these would be a worker
-            // escalating into the orchestrator role.
+            // Fleet lifecycle is the primary's alone — spawn, stop, drain. An architect and a
+            // collaborator deliberately do NOT get these, so neither can tear down or stand up
+            // workers even though one of them coordinates them; and a worker driving any of these
+            // would be a worker escalating into the orchestrator role.
             case SPAWN, STOP, DRAIN, HANDOVER -> caller.isPrimary();
 
-            // Delivering a turn is open to the primary and the architect: an architect delegates
-            // to workers (that is the role's point) but still has no lifecycle rights. A worker is
-            // excluded — sending would be it escalating.
-            case SEND -> caller.isPrimary() || caller.isArchitect();
+            // Delivering a turn to a local session is open to the primary and the architect
+            // unconditionally. A collaborator may reach only a target that is itself a configured
+            // lead or collaborator, never a spawned member's terminal. An observer may reach only
+            // a target that would itself resolve as an observer, never a lead, a collaborator, or
+            // a spawned member. A worker is excluded from every case — sending would be it
+            // escalating into the orchestrator role.
+            case SEND -> caller.isPrimary() || caller.isArchitect()
+                    || (caller.isCollaborator() && knownLeadOrCollaborator.test(targetSession))
+                    || (caller.isObserver() && knownObserverTarget.test(targetSession));
+
+            // Resolving a worker's blocked question is part of delegating to it, open to the same
+            // two roles that may stand up that delegation in the first place. Not a collaborator:
+            // resuming another session's turn is lifecycle-adjacent, not peer messaging.
+            case ANSWER -> caller.isPrimary() || caller.isArchitect();
+
+            // Leaves the daemon over the coordination broker rather than addressing a local
+            // session, open to the same two roles as ANSWER. Not a collaborator: it is a
+            // local-tab peer with no cross-host route.
+            case COORD_SEND -> caller.isPrimary() || caller.isArchitect();
 
             // The load-bearing rule: a caller acts only as the pane it occupies. CB-532 widened who
             // that can be — a lead answering another lead is replying for its OWN terminal, which
             // this already permits — while the rule itself is unchanged, and is what stops anyone
-            // forging a reply for a rendezvous someone else is waiting on. An architect's own pane
-            // passes through the same check, so it can answer a funnel that delegated to it. An
-            // unnamed primary (token/loopback, no pane) owns nothing and is still excluded.
+            // forging a reply for a rendezvous someone else is waiting on. An architect's or a
+            // collaborator's own pane passes through the same check, so each can answer a funnel
+            // that delegated to it. An unnamed primary (token/loopback, no pane) owns nothing and
+            // is still excluded.
             case REPLY, ASK -> caller.ownsSession(targetSession);
 
-            // Observation is open to every authenticated role: a worker legitimately polls its own
-            // status, and the roster carries no secrets.
-            case READ, METRICS -> caller.isPrimary() || caller.isWorker() || caller.isArchitect();
+            // READ is roster, profile, and identity observation — fleet_list, fleet_profiles, and
+            // fleet_whoami — and carries no secrets: no ticket reply, no pending question, and no
+            // other session's turn state. Those live under TASK_READ. METRICS is the separate
+            // Prometheus scrape. Both are open to every authenticated role, including a
+            // collaborator and the unconfigured-pane floor.
+            case READ, METRICS -> caller.isPrimary() || caller.isWorker() || caller.isArchitect()
+                    || caller.isCollaborator() || caller.isObserver();
+
+            // Ticket polling and session status, open unconditionally to every role READ is open
+            // to except a collaborator or an observer. A collaborator or an observer gets it too,
+            // but only for a ticket it created itself: fleet_send(wait:false) is open to both
+            // roles, so either can hold a ticket nobody else may read, and the ticket already
+            // records who created it. MessageService makes the same comparison again on every
+            // read, so this gate narrows WHO gets to ask, not what the answer is.
+            case TASK_READ -> caller.isPrimary() || caller.isWorker() || caller.isArchitect()
+                    || callerOwnsTicket.test(targetSession);
 
             // fleetd #421: reading held lead-to-lead mail is the primary's alone. An architect
             // holds READ today (CB-548), so "not primary" must mean not-architect here too — this
-            // is coordination between leads, not observation of the roster.
+            // is coordination between leads, not observation of the roster. The same reasoning
+            // excludes a collaborator.
             case COORD_READ -> caller.isPrimary();
         };
     }

fleetd/src/main/java/dev/ltms/fleet/auth/CallerResolver.java

@@ -7,6 +7,7 @@ import java.nio.charset.StandardCharsets;
 import java.security.MessageDigest;
 import java.util.Map;
 import java.util.function.Function;
+import java.util.function.Predicate;
 import java.util.function.Supplier;
 
 /**
@@ -19,6 +20,13 @@ import java.util.function.Supplier;
  *
  * <p><strong>Resolution order</strong> — connection identity first, token second, nothing third:
  * <ol>
+ *   <li>A loopback peer PID that maps to a pane this gateway itself spawned ⇒ that member's own
+ *       role: {@link Role#WORKER} for a dev, hunter, or reviewer; {@link Role#ARCHITECT} for an
+ *       architect, but only while the live slot role still confirms it (fleetd #424 — a slot
+ *       revoked from config demotes an already-bound session on its very next request, so the
+ *       roster's own role is never granted on its word alone). No tab map is consulted — a live
+ *       spawned member's identity comes from the registry that spawned it, never from a label a
+ *       pane could also carry.</li>
  *   <li>A loopback peer PID that maps to a pane named by {@code leaders:}, by the legacy
  *       {@code primary.terminal} pin, or by an operator-labelled lead tab (CB-307, CB-530, CB-531)
  *       ⇒ {@link Role#PRIMARY}, carrying that lead's
@@ -28,9 +36,11 @@ import java.util.function.Supplier;
  *       so two leads can work as peers rather than one being demoted.</li>
  *   <li>A loopback peer PID that maps to a pane bound to a CB-548 architect slot ⇒
  *       {@link Role#ARCHITECT}, carrying the slot name. Just unforgeable as a worker's, and
- *       resolved from the <em>live</em> terminal→slot binding (never a request argument), before
- *       the generic worker fallback.</li>
- *   <li>A loopback peer PID that maps to any other herdr pane ⇒ {@link Role#WORKER}. This is
+ *       resolved from the <em>live</em> terminal→slot binding (never a request argument). This is
+ *       the case the previous step does not catch: a binding with no live spawned-member session.</li>
+ *   <li>A loopback peer PID that maps to an operator-labelled collaborator tab ⇒
+ *       {@link Role#COLLABORATOR}, carrying that collaborator's name.</li>
+ *   <li>A loopback peer PID that maps to any other herdr pane ⇒ {@link Role#OBSERVER}. This is
  *       unforgeable (the OS reports the PID, herdr owns the PID→pane map) and is honoured
  *       regardless of auth mode, so enabling auth never breaks the fleet.</li>
  *   <li>Otherwise, under {@code token} mode, a valid bearer token ⇒ {@link Role#PRIMARY}.</li>
@@ -64,6 +74,20 @@ public final class CallerResolver {
     private final Supplier<Map<String, String>> architectTerminals;
     private final Function<String, MemberRole> memberSlotRoles;
     private final Function<String, String> memberSlotNames;
+    /**
+     * terminal_id → the role of the live spawned member occupying it, or {@code null} for a
+     * terminal no spawned member occupies. Consulted first, ahead of every tab map: a live
+     * spawned member's identity is its own, whatever a tab map says about the same terminal.
+     * A function rather than the roster itself, so a resolve on the hot path never scans a list —
+     * the lookup strategy is the caller's to choose.
+     */
+    private final Function<String, MemberRole> spawnedMemberRole;
+    /**
+     * terminal_id → collaborator name; empty when none are configured. A supplier for the same
+     * reason as {@link #leadTerminals}: a collaborator tab recognised after construction (the tab
+     * scan discovering a newly-labelled tab) takes effect without a restart.
+     */
+    private final Supplier<Map<String, String>> collaboratorTerminals;
 
     /** Loopback-trust resolver: no token required, historical behaviour. Test-only. */
     CallerResolver(ConnectionIdentity identity) {
@@ -124,17 +148,42 @@ public final class CallerResolver {
     /**
      * Live registry form that can confirm a bound slot is an architect slot.
      *
-     * <p>This is the only public construction path. It keeps terminal bindings and slot roles in
-     * the same {@link MemberRegistry}, so a configured architect can resolve as an architect.
+     * <p>It keeps terminal bindings and slot roles in the same {@link MemberRegistry}, so a
+     * configured architect can resolve as an architect. No spawned-member roster or collaborator
+     * registry is consulted — equivalent to {@link #withLeadsAndMembers(ConnectionIdentity,
+     * boolean, String, Supplier, MemberRegistry, Function, Supplier)} with both absent. Kept for
+     * every caller that has neither to offer, so adding them did not churn every construction site.
      */
     public static CallerResolver withLeadsAndMembers(ConnectionIdentity identity,
                                                        boolean tokenMode, String token,
                                                        Supplier<Map<String, String>> leadTerminals,
                                                        MemberRegistry members) {
+        return withLeadsAndMembers(identity, tokenMode, token, leadTerminals, members, null, null);
+    }
+
+    /**
+     * Live registry form that also resolves a live spawned member to its own role, and a
+     * configured collaborator tab to {@link Role#COLLABORATOR}.
+     *
+     * <p>This is the only public construction path that exercises the full resolution order.
+     *
+     * @param spawnedMemberRole      terminal_id → the role of the live spawned member occupying
+     *                               it, or {@code null} for a terminal no spawned member occupies.
+     *                               {@code null} here means no roster is consulted at all (every
+     *                               terminal falls through to the tab maps), not that none matches.
+     * @param collaboratorTerminals  terminal_id → collaborator name, live like {@code leadTerminals}
+     */
+    public static CallerResolver withLeadsAndMembers(ConnectionIdentity identity,
+                                                       boolean tokenMode, String token,
+                                                       Supplier<Map<String, String>> leadTerminals,
+                                                       MemberRegistry members,
+                                                       Function<String, MemberRole> spawnedMemberRole,
+                                                       Supplier<Map<String, String>> collaboratorTerminals) {
         return new CallerResolver(identity, tokenMode, token, leadTerminals,
                 members == null ? null : members::snapshot,
                 members == null ? null : members::roleForSlot,
-                members == null ? null : members::nameForSlot);
+                members == null ? null : members::nameForSlot,
+                spawnedMemberRole, collaboratorTerminals);
     }
 
     private static Supplier<Map<String, String>> fixed(Map<String, String> leadTerminals) {
@@ -160,6 +209,17 @@ public final class CallerResolver {
                            Supplier<Map<String, String>> architectTerminals,
                            Function<String, MemberRole> memberSlotRoles,
                            Function<String, String> memberSlotNames) {
+        this(identity, tokenMode, token, leadTerminals, architectTerminals, memberSlotRoles,
+                memberSlotNames, null, null);
+    }
+
+    private CallerResolver(ConnectionIdentity identity, boolean tokenMode, String token,
+                           Supplier<Map<String, String>> leadTerminals,
+                           Supplier<Map<String, String>> architectTerminals,
+                           Function<String, MemberRole> memberSlotRoles,
+                           Function<String, String> memberSlotNames,
+                           Function<String, MemberRole> spawnedMemberRole,
+                           Supplier<Map<String, String>> collaboratorTerminals) {
         if (tokenMode && (token == null || token.isBlank())) {
             throw new IllegalArgumentException(
                     "auth.mode=token requires a non-empty token; check that the env var named by "
@@ -172,6 +232,8 @@ public final class CallerResolver {
         this.architectTerminals = architectTerminals == null ? Map::of : architectTerminals;
         this.memberSlotRoles = memberSlotRoles == null ? _ -> null : memberSlotRoles;
         this.memberSlotNames = memberSlotNames == null ? Function.identity() : memberSlotNames;
+        this.spawnedMemberRole = spawnedMemberRole == null ? _ -> null : spawnedMemberRole;
+        this.collaboratorTerminals = collaboratorTerminals == null ? Map::of : collaboratorTerminals;
     }
 
     /**
@@ -188,14 +250,109 @@ public final class CallerResolver {
     }
 
     /**
-     * The currently-recognised architect slots, {@code terminal_id → slot name} (CB-548).
+     * The currently-recognised collaborator tabs, {@code terminal_id → name}.
      *
-     * <p>Read from the same supplier {@link #resolve} consults, so a slot that is <em>listed</em>
-     * here but would not <em>resolve</em> (or the reverse) cannot drift apart. Live for the same
-     * reason as {@link #leads()}.
+     * <p>Read from the same supplier {@link #resolve} consults, for the reason given in
+     * {@link #leads()}. Live for the same reason as {@link #leads()}.
      */
-    public Map<String, String> members() {
-        return architectTerminals.get();
+    public Map<String, String> collaborators() {
+        return collaboratorTerminals.get();
+    }
+
+    /**
+     * Whether {@code target} names a terminal this resolver would resolve as a lead or a
+     * collaborator — the classifier a collaborator's {@code SEND} is checked against, read from the
+     * exact maps {@link #resolve} consults so a target that would resolve as a lead or collaborator
+     * is never the one a collaborator is refused to reach, or the reverse.
+     */
+    public Predicate<String> knownLeadOrCollaborator() {
+        return target -> leadTerminals.get().containsKey(target)
+                || collaboratorTerminals.get().containsKey(target);
+    }
+
+    /**
+     * Whether {@code target} names a terminal this resolver would itself resolve as {@link
+     * Role#OBSERVER} — the classifier an observer's {@code SEND} is checked against, read from the
+     * same maps and functions {@link #resolve} consults so a target this accepts is exactly one
+     * {@code resolve} would hand back {@link Role#OBSERVER} for, and the reverse.
+     */
+    public Predicate<String> sendableObserverTarget() {
+        return target -> target != null
+                && spawnedMemberRole.apply(target) == null
+                && !leadTerminals.get().containsKey(target)
+                && !boundToArchitectSlot(target)
+                && !collaboratorTerminals.get().containsKey(target);
+    }
+
+    /**
+     * Whether {@code terminal} is bound to a configured slot the live roster still confirms as an
+     * architect — the one classifier {@link #sendableObserverTarget()} and {@code FleetMcp}'s
+     * {@code panes} row both read, so a pane's reported role and its {@code SEND} reachability can
+     * never drift apart.
+     */
+    public boolean boundToArchitectSlot(String terminal) {
+        String slot = architectTerminals.get().get(terminal);
+        return slot != null && memberSlotRoles.apply(slot) == MemberRole.ARCHITECT;
+    }
+
+    /**
+     * Resolve the Principal a bare terminal would carry, with no connection to read a pid from —
+     * for a caller deciding whether a session it already knows the terminal of, but is not itself
+     * handling a request from, may still perform some action. Applies the same precedence {@link
+     * #resolve} applies to a connection's terminal, so a terminal this method calls a collaborator
+     * or an observer is exactly one {@link #resolve} would resolve the same way.
+     *
+     * @param terminal the herdr {@code terminal_id} to resolve, or {@code null}
+     */
+    public Principal resolveTerminal(String terminal) {
+        return terminal == null ? Principal.anonymous() : resolveForTerminal(terminal, -1);
+    }
+
+    private Principal resolveForTerminal(String terminal, long pid) {
+        MemberRole spawnedRole = spawnedMemberRole.apply(terminal);
+        if (spawnedRole != null) {
+            // A live spawned member occupies this pane. Its identity is its own, whatever a tab
+            // map says about the same terminal — checked before every tab map, consulting none
+            // of them, so a tab label can never override a roster entry for the same terminal.
+            if (spawnedRole == MemberRole.ARCHITECT) {
+                // The roster only answers THAT this pane is a live spawned member; config still
+                // decides WHAT that member's slot grants (fleetd #424). A slot revoked after the
+                // bind must still demote this session on its very next request, so the roster's
+                // own ARCHITECT role is confirmed against the live slot role, exactly as the
+                // architect-slot step below confirms a binding with no live member session.
+                String slot = architectTerminals.get().get(terminal);
+                if (slot != null && memberSlotRoles.apply(slot) == MemberRole.ARCHITECT) {
+                    return Principal.architect(memberSlotNames.apply(slot), terminal, pid);
+                }
+                return Principal.worker(terminal, pid);
+            }
+            return Principal.worker(terminal, pid);
+        }
+        String lead = leadTerminals.get().get(terminal);
+        if (lead != null) {
+            // The config names this pane as a lead's own. The pane mapping is exactly as
+            // unforgeable as a worker's, so it outranks the token path — no credential needed.
+            // Checked before the architect registry so a pane named in BOTH is still the lead
+            // (CB-548 preserves every existing leader behaviour).
+            return Principal.leader(lead, terminal, pid);
+        }
+        String slot = architectTerminals.get().get(terminal);
+        if (slot != null && memberSlotRoles.apply(slot) == MemberRole.ARCHITECT) {
+            // The config/live binding names this pane as an architect slot's own. Same
+            // unforgeable pane mapping; the live binding, never a request argument, decides.
+            // Check the slot role too: this defence in depth prevents a bad lifecycle bind from
+            // escalating a dev, hunter or reviewer into an architect. This is the case the
+            // spawned-member step above does not catch: a binding with no live member session.
+            return Principal.architect(memberSlotNames.apply(slot), terminal, pid);
+        }
+        String collaborator = collaboratorTerminals.get().get(terminal);
+        if (collaborator != null) {
+            // An operator-labelled collaborator tab, confirmed live by the same scan that
+            // confirms a lead tab. Checked last among the tab maps so a pane also matching one
+            // of the above keeps that stronger role.
+            return Principal.collaborator(collaborator, terminal, pid);
+        }
+        return Principal.observer(terminal, pid); // unforgeable; never token-gated
     }
 
     /**
@@ -208,24 +365,7 @@ public final class CallerResolver {
     public Principal resolve(String remoteAddr, int remotePort, String authorizationHeader) {
         ConnectionIdentity.Caller c = identity.resolve(remoteAddr, remotePort);
         if (c.terminal() != null) {
-            String lead = leadTerminals.get().get(c.terminal());
-            if (lead != null) {
-                // The config names this pane as a lead's own. The pane mapping is exactly as
-                // unforgeable as a worker's, so it outranks the token path — no credential needed.
-                // Checked before the architect registry so a pane named in BOTH is still the lead
-                // (CB-548 preserves every existing leader behaviour).
-                return Principal.leader(lead, c.terminal(), c.pid());
-            }
-            String slot = architectTerminals.get().get(c.terminal());
-            if (slot != null && memberSlotRoles.apply(slot) == MemberRole.ARCHITECT) {
-                // The config/live binding names this pane as an architect slot's own. Same
-                // unforgeable pane mapping; the live binding, never a request argument, decides.
-                // Check the slot role too: this defence in depth prevents a bad lifecycle bind from
-                // escalating a dev, hunter or reviewer into an architect. Checked before
-                // the worker fallback.
-                return Principal.architect(memberSlotNames.apply(slot), c.terminal(), c.pid());
-            }
-            return Principal.worker(c.terminal(), c.pid()); // unforgeable; never token-gated
+            return resolveForTerminal(c.terminal(), c.pid());
         }
 
         if (tokenMode) {

fleetd/src/main/java/dev/ltms/fleet/auth/Principal.java

@@ -11,7 +11,9 @@ package dev.ltms.fleet.auth;
  * @param pid      the connecting process id, or {@code -1} when not resolvable (audit context)
  * @param name     for a lead resolved from the CB-530 {@code leaders:} registry, which lead it is;
  *                 for an architect resolved from the CB-548 {@code architects:} registry, which
- *                 slot it occupies; {@code null} for every other caller, including an unnamed primary
+ *                 slot it occupies; for a collaborator resolved from the {@code collaborators:}
+ *                 registry, which collaborator it is; {@code null} for every other caller,
+ *                 including an unnamed primary
  */
 public record Principal(Role role, String terminal, long pid, String name) {
 
@@ -73,6 +75,26 @@ public record Principal(Role role, String terminal, long pid, String name) {
         return new Principal(Role.ARCHITECT, terminal, pid, slotName);
     }
 
+    /**
+     * A collaborator: a human-opened tab recognised by its exact label in the
+     * {@code collaborators:} registry.
+     *
+     * <p>Carries {@link Role#COLLABORATOR}. {@code name} is reporting only — it lets
+     * {@code fleet_whoami} say which collaborator is asking. Identity is the {@code terminal}:
+     * like a worker's it comes from the connection, so {@code ownsSession} works exactly as it
+     * does for a worker — a collaborator acts as its own pane and no other.
+     */
+    public static Principal collaborator(String name, String terminal, long pid) {
+        return new Principal(Role.COLLABORATOR, terminal, pid, name);
+    }
+
+    /**
+     * The unconfigured-pane floor: a loopback caller whose pane matched no other role.
+     */
+    public static Principal observer(String terminal, long pid) {
+        return new Principal(Role.OBSERVER, terminal, pid);
+    }
+
     public boolean isPrimary() {
         return role == Role.PRIMARY;
     }
@@ -81,15 +103,23 @@ public record Principal(Role role, String terminal, long pid, String name) {
         return role == Role.ARCHITECT;
     }
 
+    public boolean isCollaborator() {
+        return role == Role.COLLABORATOR;
+    }
+
     public boolean isWorker() {
         return role == Role.WORKER;
     }
 
+    public boolean isObserver() {
+        return role == Role.OBSERVER;
+    }
+
     /**
      * Whether this caller is a spawned member with its own pane.
      *
-     * <p>Both workers and architects are spawned members. A lead is excluded because recording it
-     * as present would count it as an available member in the roster.
+     * <p>Both workers and architects are spawned members. A lead is not: it is a peer the
+     * operator started and named, never a pane this daemon spawned.
      */
     public boolean isSpawnedMember() {
         return role == Role.WORKER || role == Role.ARCHITECT;
@@ -115,11 +145,33 @@ public record Principal(Role role, String terminal, long pid, String name) {
         return terminal != null && terminal.equals(sessionId);
     }
 
+    /**
+     * Stable identity used to own tickets and open turns. The unnamed primary has no owner key —
+     * {@code null} — and that is matched against a ticket's recorded owner the same way any other
+     * key is: it owns a ticket another unnamed primary created, and nothing else.
+     */
+    public String ownerKey() {
+        return switch (role) {
+            case PRIMARY -> name == null ? null : prefixed("leader", name);
+            case WORKER -> prefixed("worker", terminal);
+            case ARCHITECT -> prefixed("architect", terminal);
+            case COLLABORATOR -> prefixed("collaborator", name);
+            case OBSERVER -> prefixed("observer", terminal);
+            case ANONYMOUS -> "anonymous";
+        };
+    }
+
+    private static String prefixed(String role, String identity) {
+        return role + ":" + identity;
+    }
+
     /** Short, non-sensitive description for audit lines and error details. */
     public String describe() {
         return switch (role) {
             case WORKER -> "worker:" + terminal;
             case ARCHITECT -> "architect:" + name;
+            case COLLABORATOR -> "collaborator:" + name;
+            case OBSERVER -> "observer:" + terminal;
             case PRIMARY -> name == null ? "primary" : "leader:" + name;
             case ANONYMOUS -> "anonymous";
         };

fleetd/src/main/java/dev/ltms/fleet/auth/Role.java

@@ -12,9 +12,10 @@ package dev.ltms.fleet.auth;
 public enum Role {
 
     /**
-     * The orchestrating session. Established either by being a loopback caller that is not a
-     * worker pane (under {@code loopback-trust}) or by presenting a valid bearer token (under
-     * {@code token} mode).
+     * The orchestrating session. Established either by being a loopback caller that resolves to
+     * no herdr pane at all (under {@code loopback-trust}) or by presenting a valid bearer token
+     * (under {@code token} mode). A loopback caller that does own a pane, but matches none of the
+     * roles below, resolves to {@link #OBSERVER} instead.
      */
     PRIMARY,
 
@@ -34,6 +35,29 @@ public enum Role {
      */
     ARCHITECT,
 
+    /**
+     * A config-declared, human-opened tab recognised by its exact label (the {@code
+     * fleet.collaborators.<name>.tab} registry). Never spawned — identity comes from the
+     * connection, never a request argument, exactly like {@link #WORKER} and {@link #ARCHITECT}.
+     * May {@code SEND} only to a configured lead or collaborator, {@code REPLY}/{@code ASK} only
+     * as its own pane, and {@code READ}/{@code METRICS}; may not {@code SPAWN}/{@code STOP}/
+     * {@code DRAIN}/{@code HANDOVER}, poll a ticket ({@code TASK_READ}), or reach the
+     * coordination broker ({@code COORD_SEND}/{@code COORD_READ}).
+     */
+    COLLABORATOR,
+
+    /**
+     * A loopback pane that resolved to none of the roles above: not a live spawned member, not a
+     * configured lead, not a bound architect slot, not a configured collaborator tab. Unforgeable
+     * like a worker's — derived from the connection's pane, never from a request argument, and
+     * honoured regardless of auth mode. May {@code READ} and {@code METRICS}, {@code REPLY}/
+     * {@code ASK} only as its own pane, and {@code SEND} only to a target that would itself
+     * resolve as {@code OBSERVER}; may not {@code SPAWN}/{@code STOP}/{@code DRAIN}/
+     * {@code HANDOVER}, poll a ticket ({@code TASK_READ}), or reach the coordination broker
+     * ({@code COORD_SEND}/{@code COORD_READ}).
+     */
+    OBSERVER,
+
     /** Authenticated as nothing. Authorized for nothing but {@code /healthz}. */
     ANONYMOUS
 }

fleetd/src/main/java/dev/ltms/fleet/config/ConfigRef.java

@@ -67,7 +67,7 @@ import java.util.function.Supplier;
  *       {@code models:} above: {@code dev.ltms.fleet.lead.LeadRollover} holds a
  *       {@code Supplier<FleetConfig.LeadRollover>} (the same {@code () -> config.get().x()} shape)
  *       and reads {@code handoverPath}/{@code requireOperatorConfirm}/{@code maxDocAgeSeconds}/
- *       {@code turnSettleSeconds}/{@code clearSettleSeconds}/{@code bootstrapText} fresh on every
+ *       {@code turnSettleSeconds}/{@code relaunchReadySeconds}/{@code bootstrapText} fresh on every
  *       {@code open()}/{@code confirm()} call (and on the deferred post-{@code confirm()}
  *       continuation fleetd #480's correction added — see {@code LeadRollover}'s class doc) rather
  *       than capturing them into fields at construction — unlike its closest
@@ -615,7 +615,7 @@ public final class ConfigRef implements Supplier<FleetConfig> {
         // may bind to, AND what a slot already bound still grants) through its own instance of that
         // same supplier shape — see MemberRegistry.live and its class doc for the binding rule:
         // removing a slot revokes ARCHITECT on the bound pane's very next request, and only the slot
-        // OCCUPANCY survives, so the demoted session keeps its slot key until it unbinds. Only
+        // OCCUPANCY survives, so the demoted session keeps its slot key until it unbinds.
         // fleet.leaders is frozen (Fleetd.java:281 reads cfg.fleet().leaders() off the startup
         // snapshot to build both the LeadTabScanner's tab-label-to-name map, wired into
         // CallerResolver.withLeadsAndMembers at Fleetd.java:620/624, and — when herdr answered —
@@ -635,6 +635,11 @@ public final class ConfigRef implements Supplier<FleetConfig> {
                     + "live through that same supplier for placement AND through a separate supplier "
                     + "on MemberRegistry for spawn-time identity — both already applied");
         }
+        if (!Objects.equals(collaboratorsOf(old), collaboratorsOf(fresh))) {
+            changed.add("fleet: fleet.collaborators (each collaborator's tab) is read once at "
+                    + "startup to build the LeadTabScanner's identity map, which is not rebuilt on "
+                    + "reload, so a collaborator added, removed, or given a new tab: needs a restart");
+        }
         // Kept in step with SPLIT_KEYS the same way changedColdKeys is kept in step with COLD_KEYS —
         // every message here must be traceable to one of the split keys the class doc documents.
         // NOTE what this does NOT prove, per the javadoc above: it does not catch a SPLIT_KEYS
@@ -650,6 +655,11 @@ public final class ConfigRef implements Supplier<FleetConfig> {
         return cfg.fleet() == null ? Map.of() : cfg.fleet().leaders();
     }
 
+    /** {@code cfg.fleet().collaborators()}, defensively, in case a caller hands in a non-defaulted config. */
+    private static Map<String, FleetConfig.Collaborator> collaboratorsOf(FleetConfig cfg) {
+        return cfg.fleet() == null ? Map.of() : cfg.fleet().collaborators();
+    }
+
     /**
      * {@link FleetConfig.Profile} record components deliberately left out of
      * {@link #sameLaunchSettings} because they are read <em>live</em>, not baked in at spawn — see

fleetd/src/main/java/dev/ltms/fleet/config/FleetConfig.java

@@ -28,6 +28,7 @@ import java.util.Comparator;
 import java.util.HashSet;
 import java.util.LinkedHashMap;
 import java.util.List;
+import java.util.Locale;
 import java.util.Map;
 import java.util.Set;
 import java.util.regex.Pattern;
@@ -1092,8 +1093,8 @@ public record FleetConfig(
     }
 
     /**
-     * One entry of the CB-530 {@code leaders:} registry — a pane that orchestrates rather than one
-     * that is orchestrated.
+     * One entry of the {@code leaders:} registry — a pane that orchestrates rather than one that is
+     * orchestrated.
      *
      * <p>Why a registry and not a second {@code primary:}: {@code primary.terminal} is singular by
      * construction, so a session in any other pane resolves as a worker. That is correct while one
@@ -1103,49 +1104,49 @@ public record FleetConfig(
      * <p>{@code kind} and {@code model} are descriptive only: they document what runs in the pane
      * and are reported back by {@code fleet_whoami}.
      *
-     * <p><b>A lead is now also creatable (CB-557).</b> Before, nothing spawned one — a lead
-     * pre-existed, which is why it had to be recognised by configuration rather than created. With
-     * {@code profile} and {@code instances} the daemon may stand one up when none is live, so the
-     * pane no longer has to exist before the daemon does. Recognition still comes first: a lead
-     * already running in its configured {@code tab} is adopted, and only the shortfall is launched.
+     * <p>A lead with {@code profile} and {@code instances} set may be launched by the daemon when
+     * none is live; recognition always comes first, so only the shortfall is launched.
      *
-     * <p><b>{@code tab} replaced {@code terminal} (CB-579).</b> A herdr {@code terminal_id} changes
-     * every time the lead's session restarts, so pinning one cost a config edit and a daemon restart
-     * per restart. A tab is stable: a human opens it once, it holds exactly one pane, and its label
-     * survives restarts of the agent inside it — so identity is now the tab label alone.
+     * <p>Every lead's tab is labelled {@link #LEAD_TAB_LABEL}, a fixed constant — not a per-entry
+     * config value. {@code workspace} is therefore what tells one lead from another: two leaders
+     * sharing one space would both resolve to the one tab named {@code lead} there, so only one
+     * could ever be found. {@code tab} is a deprecated legacy label, still matched within this
+     * lead's own space alongside the constant.
      *
      * @param profile             the {@code profiles:} entry to launch this lead on when one must
      *                            be created; {@code null} ⇒ recognise-only, never create.
-     *                            <p>fleetd #176: also the field {@code Fleetd.leadSeatLookup} reads
-     *                            to learn which account this lead's own live session shares — set it
+     *                            <p>Also the field {@code Fleetd.leadSeatLookup} reads to learn
+     *                            which account this lead's own live session shares — set it
      *                            (safely, even on an already-running recognise-only lead: naming a
      *                            profile here never starts anything beyond {@code instances}) so a
      *                            {@code subscription: true} worker profile sharing its
      *                            {@code effectiveCredentialId()} has this lead's seat subtracted from
      *                            {@code fleet_list}'s {@code free}. {@code null} here also means this
      *                            lead's seat cannot be derived and is not counted.
-     * @param tab                 the exact tab label hosting this lead, matched case-insensitively;
-     *                            the only field identity depends on. Required — a lead with no
-     *                            {@code tab} can never be discovered, launched or not
+     * @param tab                 deprecated legacy tab label, matched case-insensitively within
+     *                            this lead's own space alongside {@link #LEAD_TAB_LABEL}. Optional —
+     *                            {@code null}/blank means only the constant is accepted
      * @param instances           how many of this lead should be live (default 1). The daemon
      *                            launches only the shortfall, so a restart adopts rather than doubles
-     * @param tabPrefix           no longer used to find a lead's tab — {@code tab} is matched
-     *                            exactly. Its only remaining job is the startup collision guard
-     *                            ({@link #validateLeadTabPrefixes()}), which still uses it to refuse
-     *                            a worker {@code tabLabel} template that could be misread as a lead.
-     *                            Default {@code "lead:"}
+     * @param tabPrefix           lead-tab naming convention checked against member labels. Lead
+     *                            identity uses {@link #acceptedLabels()}. Default {@code "lead:"}
      * @param scanIntervalSeconds how long a tab scan is cached before herdr is asked again; also the
      *                            worst case before a newly-labelled tab is recognised. Default 10
      * @param kind                which agent runs there ({@code claude}, {@code opencode}, …)
      * @param model               the model or selector it runs, for operators reading the roster
+     * @param workspace           the space this lead's tab lives in — the uniqueness boundary
+     *                            identity now depends on. Default {@link #DEFAULT_WORKSPACE}
      */
     @JsonIgnoreProperties(ignoreUnknown = true)
     public record Leader(String profile, String tab, Integer instances, String tabPrefix,
                          Integer scanIntervalSeconds, String kind, String model,
                          String workspace, String cwd) {
 
+        /** The tab label every lead is found by, and an auto-launched instance is created with. */
+        public static final String LEAD_TAB_LABEL = "lead";
+
         /**
-         * Where an auto-launched lead's tab is created (CB-558). It defaults to the SAME shared
+         * Where an auto-launched lead's tab is created. It defaults to the SAME shared
          * {@code "fleet"} space the members use, so the operator sees one "session" with many tabs.
          * The scanner no longer excludes member spaces — it tells a lead from a member by the exact
          * tab label, so a lead sharing the members' space is still discovered (see LeadLauncher).
@@ -1173,9 +1174,42 @@ public record FleetConfig(
             return profile != null && !profile.isBlank() && instances > 0;
         }
 
-        /** The tab label an auto-launched instance of this lead gets — its configured {@code tab}. */
+        /** The tab label an auto-launched instance of this lead gets. */
         public String tabLabel() {
-            return tab;
+            return LEAD_TAB_LABEL;
+        }
+
+        /**
+         * The normalised labels (stripped, lower-cased) a tab in this lead's own space may carry to
+         * be recognised as this lead: {@link #LEAD_TAB_LABEL} first, plus the deprecated {@code tab}
+         * when configured and different. Every matcher in this class's callers reads this method —
+         * none re-derives the set.
+         */
+        public List<String> acceptedLabels() {
+            String normalizedTab = (tab == null) ? null : tab.toLowerCase(Locale.ROOT);
+            if (normalizedTab == null || normalizedTab.equals(LEAD_TAB_LABEL)) {
+                return List.of(LEAD_TAB_LABEL);
+            }
+            return List.of(LEAD_TAB_LABEL, normalizedTab);
+        }
+    }
+
+    /**
+     * A tab fleetd recognises as a collaborator, keyed by name (fleetd #669).
+     *
+     * <p>Recognise-only: there is no {@code profile}, no {@code instances} and no {@code kind}.
+     * Nothing here ever launches a pane.
+     *
+     * <p>{@code tabPrefix} is absent. Identity is matched on the exact {@code tab} alone.
+     *
+     * @param tab the exact tab label hosting this collaborator, matched case-insensitively; the
+     *            only field identity depends on. Required — an entry with no {@code tab} can
+     *            never be discovered.
+     */
+    @JsonIgnoreProperties(ignoreUnknown = true)
+    public record Collaborator(String tab) {
+        public Collaborator {
+            tab = (tab == null || tab.isBlank()) ? null : tab.strip();
         }
     }
 
@@ -1216,15 +1250,18 @@ public record FleetConfig(
      * is exactly compatible with that. The pool is also what replaced {@code defaultProfile:} — an
      * unqualified spawn names a role, and the role's pool supplies the candidates.
      *
-     * @param leaders     panes that orchestrate rather than are orchestrated, keyed by lead name
-     * @param architects  profiles the {@code architect} role may run on
-     * @param developers  profiles the {@code dev} role may run on
-     * @param hunters     profiles the {@code hunter} role may run on
-     * @param reviewers   profiles the {@code reviewer} role may run on
-     * @param charters    optional launch-charter text keyed by singular role wire name
-     * @param tabLabel    template for a member tab's label; {@code {role}}, {@code {profile}},
-     *                    {@code {model}} and {@code {n}} (a per role+profile counter) are
-     *                    substituted. Default {@link #DEFAULT_TAB_LABEL}
+     * @param leaders       panes that orchestrate rather than are orchestrated, keyed by lead name
+     * @param architects    profiles the {@code architect} role may run on
+     * @param developers    profiles the {@code dev} role may run on
+     * @param hunters       profiles the {@code hunter} role may run on
+     * @param reviewers     profiles the {@code reviewer} role may run on
+     * @param charters      optional launch-charter text keyed by singular role wire name
+     * @param tabLabel      template for a member tab's label; {@code {role}}, {@code {profile}},
+     *                      {@code {model}} and {@code {n}} (a per role+profile counter) are
+     *                      substituted. Default {@link #DEFAULT_TAB_LABEL}
+     * @param collaborators tabs fleetd recognises as collaborators (fleetd #669), keyed by name.
+     *                      Recognise-only, exactly like a {@code profile}-less {@link Leader}:
+     *                      nothing here is ever auto-launched.
      */
     @JsonIgnoreProperties(ignoreUnknown = true)
     public record Fleet(Map<String, Leader> leaders,
@@ -1233,13 +1270,11 @@ public record FleetConfig(
                         Map<String, Slot> hunters,
                         Map<String, Slot> reviewers,
                         Map<String, String> charters,
-                        String tabLabel) {
+                        String tabLabel,
+                        Map<String, Collaborator> collaborators) {
 
         /**
-         * Role first, so the tab bar reads as the fleet and so the label shares a namespace with a
-         * lead's {@code tabPrefix}. Because {@code {role}} comes from a closed enum, a generated
-         * member label can never begin with {@code "lead:"} — the clash that
-         * {@link #validateLeadTabPrefixes()} used to have to check for is unrepresentable here.
+         * Role first, so the tab bar identifies the member's fleet role.
          */
         public static final String DEFAULT_TAB_LABEL = "{role}: {profile} #{n}";
 
@@ -1251,26 +1286,30 @@ public record FleetConfig(
             reviewers = unmodifiableOrEmpty(reviewers);
             charters = unmodifiableOrEmpty(charters);
             tabLabel = (tabLabel == null || tabLabel.isBlank()) ? DEFAULT_TAB_LABEL : tabLabel;
+            collaborators = unmodifiableOrEmpty(collaborators);
         }
 
         /**
-         * A fleet with no configured launch charters — the shape every deployment had before
-         * CB-566, and what most tests want.
+         * A fleet with no configured launch charters and no collaborators — the shape every
+         * deployment had before CB-566, and what most tests want.
          *
          * <p>Kept deliberately, even though an overload that drops a new field is normally the
          * shape to avoid. It is safe here because nothing <em>reads</em> a charter through a
          * constructor: the launcher reads {@code fleet.charters()} from the live config. Jackson
          * binds the canonical constructor, so this one cannot swallow an operator's YAML.
+         * {@code collaborators} is dropped the same way and for the same reason: no caller of
+         * this overload has ever needed to set it, so it defaults to empty here exactly as the
+         * canonical constructor would default an absent YAML key.
          */
         public Fleet(Map<String, Leader> leaders, Map<String, Slot> architects,
                      Map<String, Slot> developers, Map<String, Slot> reviewers,
                      Map<String, String> charters, String tabLabel) {
-            this(leaders, architects, developers, null, reviewers, charters, tabLabel);
+            this(leaders, architects, developers, null, reviewers, charters, tabLabel, null);
         }
 
         public Fleet(Map<String, Leader> leaders, Map<String, Slot> architects,
                      Map<String, Slot> developers, Map<String, Slot> reviewers, String tabLabel) {
-            this(leaders, architects, developers, null, reviewers, null, tabLabel);
+            this(leaders, architects, developers, null, reviewers, null, tabLabel, null);
         }
 
         /**
@@ -1395,17 +1434,16 @@ public record FleetConfig(
      * before anything exists to call — the same fact already true of adding a brand-new
      * {@code profiles:} entry.
      *
-     * <p><strong>{@code turnSettleSeconds} (fleetd #480 correction):</strong> {@code confirm()} is
-     * called FROM the calling lead's own turn, so its pane is still {@code WORKING} the instant
-     * {@code confirm()} validates every gate and schedules the roll. {@code
-     * dev.ltms.fleet.lead.LeadRollover}'s deferred continuation waits up to this many seconds for
-     * that SAME pane to report {@code IDLE} or {@code DONE} — i.e. for the calling turn to actually
-     * end — before it sends {@code /clear} at all. {@code BLOCKED} does not count: that is a live
-     * turn merely paused, not one that has finished. If that wait times out, no {@code /clear} is
-     * ever sent: a lead that never goes idle is still doing real work, and clearing it would
-     * destroy live context. This is a separate wait from {@code clearSettleSeconds} below, which
-     * bounds the SECOND wait, for the pane to reach {@code IDLE} or {@code DONE} again AFTER
-     * {@code /clear} has already gone out.
+     * <p><strong>{@code turnSettleSeconds}:</strong> {@code confirm()} is called FROM the calling
+     * lead's own turn, so its pane is still {@code WORKING} the instant {@code confirm()} validates
+     * every gate and schedules the roll. {@code dev.ltms.fleet.lead.LeadRollover}'s deferred
+     * continuation waits up to this many seconds for that SAME pane to report {@code IDLE} or
+     * {@code DONE} — i.e. for the calling turn to actually end — before it ends the old pane's
+     * process at all. {@code BLOCKED} does not count: that is a live turn merely paused, not one
+     * that has finished. If that wait times out, the old pane is never touched: a lead that never
+     * goes idle is still doing real work, and the roll ends that pane's whole process — there is no
+     * way back from this once it runs, so this wait is the only thing standing between "still
+     * working" and "gone".
      *
      * @param handoverPath required when this block is present — where the handover file a fresh
      *                     lead session reads must live. There is no sane non-null default for an
@@ -1426,36 +1464,49 @@ public record FleetConfig(
      *                     attempt can never be mistaken for a fresh one.
      * @param turnSettleSeconds default 300 — bound on how long the deferred roll waits for the
      *                     CALLING lead's own turn to end (its pane to report {@code IDLE} or
-     *                     {@code DONE}) before sending {@code /clear} at all. See the paragraph
-     *                     above.
-     * @param clearSettleSeconds default 20 — bound on how long to wait for the lead's pane to
-     *                     report {@code IDLE} or {@code DONE} again after {@code /clear} before
-     *                     giving up. A roll that times out here never sends {@code bootstrapText}.
+     *                     {@code DONE}) before ending that pane's process at all. See the
+     *                     paragraph above.
+     * @param relaunchReadySeconds default 45 — bound on EACH of two separate waits that run after
+     *                     the old lead's pane has been torn down and a fresh one launched: first,
+     *                     for the fresh pane itself to reach a real turn boundary ({@code IDLE} or
+     *                     {@code DONE}, never merely {@code BLOCKED}) — the safety gate, since
+     *                     typing into a pane that has not finished booting loses the keystrokes;
+     *                     second, for the fresh terminal to show up as a recognised lead, which is
+     *                     bookkeeping rather than a safety gate, so a timeout on this second wait
+     *                     does not withhold {@code bootstrapText} — it is sent once the pane is
+     *                     ready regardless. Recognition comes from the same periodically-refreshed
+     *                     scan {@code LeadTabScanner} already keeps ({@code scanIntervalSeconds},
+     *                     10s live), so a budget has to clear more than one scan interval to leave
+     *                     any real margin for the CLI's own boot time; 20 was rejected for exactly
+     *                     that reason — at a 10s scan interval it only buys two scans. 45 buys
+     *                     roughly four. Only a timeout on the FIRST wait (the pane never becomes
+     *                     ready) withholds {@code bootstrapText}.
      * @param bootstrapText default a sentence naming the RESOLVED handover path — sent to the
-     *                     lead's pane once it settles after {@code /clear}, telling the fresh
-     *                     session where to read the handover and carry on. Left {@code null} here
-     *                     when the operator configures none: the default sentence cannot be built
-     *                     at construction time because it must name the path AFTER {@code
-     *                     dev.ltms.fleet.lead.LeadRollover#open} has resolved a relative {@code
-     *                     handoverPath} against the calling lead's workspace, which this record has
-     *                     no way to know — see {@link #bootstrapTextFor(String)}.
+     *                     fresh lead's pane once it reaches a real turn boundary after relaunch,
+     *                     telling the fresh session where to read the handover and carry on. Left
+     *                     {@code null} here when the operator configures none: the default sentence
+     *                     cannot be built at construction time because it must name the path AFTER
+     *                     {@code dev.ltms.fleet.lead.LeadRollover#open} has resolved a relative
+     *                     {@code handoverPath} against the calling lead's workspace, which this
+     *                     record has no way to know — see {@link #bootstrapTextFor(String)}.
      */
     @JsonIgnoreProperties(ignoreUnknown = true)
     public record LeadRollover(String handoverPath, Boolean requireOperatorConfirm,
                                 Integer maxDocAgeSeconds, Integer turnSettleSeconds,
-                                Integer clearSettleSeconds, String bootstrapText) {
+                                Integer relaunchReadySeconds, String bootstrapText) {
         public LeadRollover {
             requireOperatorConfirm = requireOperatorConfirm == null || requireOperatorConfirm;
             maxDocAgeSeconds = (maxDocAgeSeconds == null || maxDocAgeSeconds <= 0) ? 3600 : maxDocAgeSeconds;
             turnSettleSeconds = (turnSettleSeconds == null || turnSettleSeconds <= 0) ? 300 : turnSettleSeconds;
-            clearSettleSeconds = (clearSettleSeconds == null || clearSettleSeconds <= 0) ? 20 : clearSettleSeconds;
+            relaunchReadySeconds = (relaunchReadySeconds == null || relaunchReadySeconds <= 0)
+                    ? 45 : relaunchReadySeconds;
             bootstrapText = (bootstrapText == null || bootstrapText.isBlank()) ? null : bootstrapText;
         }
 
         /**
-         * The text actually sent to the lead's pane once it settles after {@code /clear}: the
-         * operator's configured {@link #bootstrapText} when one is set, otherwise the default
-         * sentence built from {@code resolvedHandoverPath}.
+         * The text actually sent to the fresh lead's pane once it reaches a real turn boundary
+         * after relaunch: the operator's configured {@link #bootstrapText} when one is set,
+         * otherwise the default sentence built from {@code resolvedHandoverPath}.
          *
          * @param resolvedHandoverPath the ABSOLUTE path {@code dev.ltms.fleet.lead.LeadRollover
          *                              #open} already resolved — never the raw configured {@link
@@ -1898,6 +1949,7 @@ public record FleetConfig(
             rejectNegativeMaxLoad(yaml);
             rejectAutoCompactWindowOutOfRange(yaml);
             warnConflictingAutoCompactWindows(yaml);
+            warnRetiredClearSettleSecondsKey(yaml);
             rejectMalformedProfilePatterns(yaml);
             rejectUnknownKind(yaml);
             rejectUnknownAuthMode(yaml);
@@ -1916,7 +1968,7 @@ public record FleetConfig(
 
     /** The {@code fleet:} child blocks whose direct children are slot names. */
     private static final Set<String> FLEET_POOL_KEYS =
-            Set.of("leaders", "architects", "developers", "hunters", "reviewers");
+            Set.of("leaders", "architects", "developers", "hunters", "reviewers", "collaborators");
 
     /**
      * Reject a {@code fleet:} role pool whose slot names repeat (CB-548, re-homed by CB-557).
@@ -1926,7 +1978,7 @@ public record FleetConfig(
      * daemon would never know. Jackson's YAML parser does not fail on duplicate mapping keys by
      * default, so duplicates are caught here, at parse time, before the map is built.
      *
-     * <p>Only the five pools <em>directly under the top-level {@code fleet:}</em> are considered,
+     * <p>Only the six pools <em>directly under the top-level {@code fleet:}</em> are considered,
      * and only their direct child keys (the slot names). A nested field elsewhere, even one also
      * named {@code developers:}, is ignored, so parsing of the rest of the config is unaffected.
      *
@@ -2061,13 +2113,6 @@ public record FleetConfig(
         }
     }
 
-    /**
-     * The top-level keys in {@code yaml} that this build does not understand, sorted. Package-private
-     * so the guardrail is asserted directly rather than through a log appender.
-     *
-     * @return empty when everything is known, or when {@code yaml} is not a mapping at all (a
-     *         malformed file is {@code readValue}'s error to report, not this method's)
-     */
     /**
      * Top-level keys renamed by the member taxonomy, mapped old → new.
      *
@@ -2321,6 +2366,33 @@ public record FleetConfig(
                 names, String.join(", ", detail));
     }
 
+    /**
+     * Warn when a {@code leadRollover:} block still sets the retired {@code clearSettleSeconds}
+     * key. {@link LeadRollover} carries {@code @JsonIgnoreProperties(ignoreUnknown = true)} and no
+     * longer declares that component, so Jackson drops it with no signal of its own — this raw-YAML
+     * check is the only place an operator's now-inert setting is reported at all; by the time a
+     * {@link LeadRollover} instance exists to run a validator against, the key is already gone.
+     *
+     * @param yaml the raw config text
+     */
+    static void warnRetiredClearSettleSecondsKey(String yaml) {
+        Map<?, ?> raw;
+        try {
+            raw = YAML.readValue(yaml, Map.class);
+        } catch (IOException | IllegalArgumentException e) {
+            return;
+        }
+        if (raw == null || !(raw.get("leadRollover") instanceof Map<?, ?> leadRollover)) {
+            return;
+        }
+        if (leadRollover.containsKey("clearSettleSeconds")) {
+            log.warn("leadRollover.clearSettleSeconds is retired and no longer read. Set "
+                    + "leadRollover.relaunchReadySeconds instead: it bounds how long to wait, after "
+                    + "a lead is relaunched, for its pane to become ready and then for it to be "
+                    + "recognised as a lead. Remove clearSettleSeconds from fleetd.yaml.");
+        }
+    }
+
     /**
      * Reject a profile whose {@code errorPattern} (fleetd #201 Unit 5) or {@code exhaustedPattern}
      * (CB-578 stage A) is not a valid Java regex, naming the profile, the key, and the parser's own
@@ -2575,6 +2647,13 @@ public record FleetConfig(
         }
     }
 
+    /**
+     * The top-level keys in {@code yaml} that this build does not understand, sorted. Package-private
+     * so the guardrail is asserted directly rather than through a log appender.
+     *
+     * @return empty when everything is known, or when {@code yaml} is not a mapping at all (a
+     *         malformed file is {@code readValue}'s error to report, not this method's)
+     */
     static List<String> unknownTopLevelKeys(String yaml) {
         Map<?, ?> raw;
         try {
@@ -2685,83 +2764,224 @@ public record FleetConfig(
     }
 
     /**
-     * Reject a lead-scan convention that a worker tab would also satisfy (CB-531).
+     * Reject a member tab-label template that could render as a configured lead or collaborator
+     * tab, as the fixed lead tab label, or that matches a lead-tab naming convention; reject two
+     * {@code fleet.leaders} entries that share one space; reject two {@code fleet.collaborators}
+     * entries — or a lead and a collaborator — that share one exact tab; and reject a collaborator
+     * tab equal to the fixed lead tab label.
      *
-     * <p>The scan reads a tab label and concludes "a lead lives here". fleetd also <em>writes</em>
-     * tab labels — every member gets one rendered into its tab. Choose a lead {@code tabPrefix} that
-     * a member template matches and the daemon starts labelling its own members as leads, promoting
-     * the entire fleet to {@link dev.ltms.fleet.auth.Role#PRIMARY} with no message and no diff.
-     * {@link #validatePanePlacementAgainstLeadTabs()} is the check that stops a pane-placed member
-     * from landing inside a lead's tab in the first place; this check is a second, independent
-     * guard that catches the hazard even when every profile places members correctly, by refusing
-     * a label that a scan would still misread as a lead.
+     * <p>{@code fleet.collaborators} has no {@code tabPrefix}: identity is matched on the exact
+     * {@code tab} alone, so only the exact-render check applies there, not the prefix check.
      *
-     * <p>CB-557 shrank this check rather than removing it. The default template is
-     * {@code "{role}: {profile} #{n}"} and {@code {role}} comes from a closed enum, so a
-     * <em>generated</em> label can no longer collide by construction. What remains checkable is what
-     * an operator still writes by hand: the {@code fleet.tabLabel} template and any per-profile
-     * {@code tabLabel} override.
-     *
-     * <p>Fatal rather than a warning, unlike {@link #warnUnknownTopLevelKeys}: an unknown key means
-     * a feature does nothing, while this means a feature does the opposite of what it says.
-     *
-     * @throws IllegalStateException when the fleet template or any profile's {@code tabLabel}
-     *                               override starts with a configured lead prefix
+     * @throws IllegalStateException when the fleet template or a profile {@code tabLabel} override
+     *                               can render as a configured lead or collaborator tab, as the
+     *                               fixed lead tab label, or match a lead-tab prefix; when two
+     *                               leaders share one space; when two collaborators (or a lead and
+     *                               a collaborator) carry the same exact {@code tab}
+     *                               (case-insensitively); or when a collaborator's {@code tab}
+     *                               equals the fixed lead tab label
      */
     public void validateLeadTabPrefixes() {
-        if (fleet == null || fleet.leaders().isEmpty()) {
+        if (fleet == null) {
             return;
         }
         List<String> bad = new ArrayList<>();
+        if (templateCanRenderAs(fleet.tabLabel(), Leader.LEAD_TAB_LABEL)) {
+            bad.add("fleet.tabLabel=\"" + fleet.tabLabel() + "\" can render as \""
+                    + Leader.LEAD_TAB_LABEL + "\", the fixed lead tab label");
+        }
+        profiles().entrySet().stream()
+                .map(Map.Entry::getKey)
+                .sorted()
+                .forEach(p -> {
+                    String label = profiles().get(p).tabLabel();
+                    if (templateCanRenderAs(label, Leader.LEAD_TAB_LABEL)) {
+                        bad.add("profile '" + p + "' overrides tabLabel with \"" + label
+                                + "\", which can render as \"" + Leader.LEAD_TAB_LABEL
+                                + "\", the fixed lead tab label");
+                    }
+                });
         fleet.leaders().forEach((leadName, leader) -> {
             if (leader == null) {
                 return;
             }
+            String tab = leader.tab();
             String prefix = leader.tabPrefix();
-            // The fleet-wide template is checked once per prefix: it labels every member that has no
-            // override, so one bad template promotes the entire fleet, not one profile.
-            if (startsWithIgnoreCase(fleet.tabLabel(), prefix)) {
+            if (templateCanRenderAs(fleet.tabLabel(), tab)) {
+                bad.add("fleet.tabLabel=\"" + fleet.tabLabel() + "\" can render as the tab of "
+                        + "lead '" + leadName + "' (\"" + tab + "\")");
+            } else if (startsWithIgnoreCase(fleet.tabLabel(), prefix)) {
                 bad.add("fleet.tabLabel=\"" + fleet.tabLabel() + "\" starts with the tabPrefix of "
                         + "lead '" + leadName + "' (\"" + prefix + "\")");
             }
             profiles().entrySet().stream()
-                    .filter(e -> startsWithIgnoreCase(e.getValue().tabLabel(), prefix))
                     .map(Map.Entry::getKey)
                     .sorted()
-                    .forEach(p -> bad.add("profile '" + p + "' overrides tabLabel with \""
-                            + profiles().get(p).tabLabel() + "\", which starts with the tabPrefix of "
-                            + "lead '" + leadName + "' (\"" + prefix + "\")"));
+                    .forEach(p -> {
+                        String label = profiles().get(p).tabLabel();
+                        if (templateCanRenderAs(label, tab)) {
+                            bad.add("profile '" + p + "' overrides tabLabel with \"" + label
+                                    + "\", which can render as the tab of lead '" + leadName
+                                    + "' (\"" + tab + "\")");
+                        } else if (startsWithIgnoreCase(label, prefix)) {
+                            bad.add("profile '" + p + "' overrides tabLabel with \"" + label
+                                    + "\", which starts with the tabPrefix of lead '" + leadName
+                                    + "' (\"" + prefix + "\")");
+                        }
+                    });
         });
-        if (bad.isEmpty()) {
+        fleet.collaborators().forEach((collabName, collaborator) -> {
+            if (collaborator == null) {
+                return;
+            }
+            String tab = collaborator.tab();
+            if (tab != null && tab.equalsIgnoreCase(Leader.LEAD_TAB_LABEL)) {
+                bad.add("fleet.collaborators." + collabName + ".tab=\"" + tab + "\" is the fixed "
+                        + "lead tab label — a collaborator there would shadow a lead");
+            }
+            if (templateCanRenderAs(fleet.tabLabel(), tab)) {
+                bad.add("fleet.tabLabel=\"" + fleet.tabLabel() + "\" can render as the tab of "
+                        + "collaborator '" + collabName + "' (\"" + tab + "\")");
+            }
+            profiles().entrySet().stream()
+                    .map(Map.Entry::getKey)
+                    .sorted()
+                    .forEach(p -> {
+                        String label = profiles().get(p).tabLabel();
+                        if (templateCanRenderAs(label, tab)) {
+                            bad.add("profile '" + p + "' overrides tabLabel with \"" + label
+                                    + "\", which can render as the tab of collaborator '"
+                                    + collabName + "' (\"" + tab + "\")");
+                        }
+                    });
+        });
+        if (!bad.isEmpty()) {
+            throw new IllegalStateException("refusing to start: " + String.join("; ", bad)
+                    + ". A member labelled that way, while its pane carries no entry in the "
+                    + "spawned-member roster, is read back as a lead or collaborator and granted "
+                    + "that identity's authority. Change one of the two so member tabs cannot be "
+                    + "confused with a lead's or collaborator's tab.");
+        }
+
+        List<String> collisions = new ArrayList<>();
+        List<String> leadNames = fleet.leaders().keySet().stream().sorted().toList();
+        for (int i = 0; i < leadNames.size(); i++) {
+            String nameA = leadNames.get(i);
+            Leader a = fleet.leaders().get(nameA);
+            if (a == null) {
+                continue;
+            }
+            for (int j = i + 1; j < leadNames.size(); j++) {
+                String nameB = leadNames.get(j);
+                Leader b = fleet.leaders().get(nameB);
+                if (b == null) {
+                    continue;
+                }
+                if (a.workspace().equalsIgnoreCase(b.workspace())) {
+                    collisions.add("lead '" + nameA + "' and lead '" + nameB + "' share workspace \""
+                            + a.workspace() + "\" — both would resolve to the tab named \""
+                            + Leader.LEAD_TAB_LABEL + "\" in that space, so only one could ever be "
+                            + "found");
+                }
+            }
+        }
+        List<String> collabNames = fleet.collaborators().keySet().stream().sorted().toList();
+        for (int i = 0; i < collabNames.size(); i++) {
+            String nameA = collabNames.get(i);
+            Collaborator a = fleet.collaborators().get(nameA);
+            if (a == null || a.tab() == null || a.tab().isBlank()) {
+                continue;
+            }
+            for (int j = i + 1; j < collabNames.size(); j++) {
+                String nameB = collabNames.get(j);
+                Collaborator b = fleet.collaborators().get(nameB);
+                if (b == null || b.tab() == null || b.tab().isBlank()) {
+                    continue;
+                }
+                if (a.tab().equalsIgnoreCase(b.tab())) {
+                    collisions.add("collaborator '" + nameA + "' and collaborator '" + nameB
+                            + "' both use tab \"" + a.tab() + "\"");
+                }
+            }
+        }
+        for (String leadName : leadNames) {
+            Leader lead = fleet.leaders().get(leadName);
+            if (lead == null || lead.tab() == null || lead.tab().isBlank()) {
+                continue;
+            }
+            for (String collabName : collabNames) {
+                Collaborator collaborator = fleet.collaborators().get(collabName);
+                if (collaborator == null || collaborator.tab() == null
+                        || collaborator.tab().isBlank()) {
+                    continue;
+                }
+                if (lead.tab().equalsIgnoreCase(collaborator.tab())) {
+                    collisions.add("lead '" + leadName + "' and collaborator '" + collabName
+                            + "' both use tab \"" + lead.tab() + "\"");
+                }
+            }
+        }
+        if (collisions.isEmpty()) {
             return;
         }
-        throw new IllegalStateException("refusing to start: " + String.join("; ", bad)
-                + ". Every member labelled that way would be read back as a lead and granted "
-                + "spawn/stop/send on the whole fleet. Change one of the two so member tabs and "
-                + "lead tabs cannot be confused.");
+        throw new IllegalStateException("refusing to start: " + String.join("; ", collisions)
+                + ". Identity is matched exactly, so only one of two entries sharing a space or a "
+                + "tab can ever be found — the other is silently unreachable. Give each lead its "
+                + "own space, and each collaborator its own exact tab.");
+    }
+
+    private static boolean templateCanRenderAs(String template, String tab) {
+        if (template == null || template.isBlank() || tab == null || tab.isBlank()) {
+            return false;
+        }
+        var placeholders = Pattern.compile("\\{(?:role|profile|model|n)}").matcher(template);
+        StringBuilder expression = new StringBuilder("^");
+        int literalStart = 0;
+        while (placeholders.find()) {
+            expression.append(Pattern.quote(template.substring(literalStart, placeholders.start())));
+            expression.append(".*");
+            literalStart = placeholders.end();
+        }
+        expression.append(Pattern.quote(template.substring(literalStart))).append("$");
+        return Pattern.compile(expression.toString(), Pattern.CASE_INSENSITIVE).matcher(tab).matches();
+    }
+
+    /** Case-insensitive prefix test that tolerates a null or blank label. */
+    private static boolean startsWithIgnoreCase(String label, String prefix) {
+        if (label == null || prefix == null || prefix.isBlank()) {
+            return false;
+        }
+        String stripped = label.strip();
+        return stripped.regionMatches(true, 0, prefix, 0, prefix.length());
     }
 
     /**
-     * Reject a profile that places its members by {@code "pane"} while any {@code fleet.leaders}
-     * entry names a {@code tab}. A pane-placed member lands inside the focused tab rather than its
-     * own, so it can land inside a lead's own labelled tab. {@link
-     * dev.ltms.fleet.herdr.LeadTabScanner} identifies a lead purely by that tab's label — it does
-     * not exclude the member space — so a member that ends up there would be read back as the lead
-     * and granted spawn/stop/send on the whole fleet.
+     * Reject a profile that places its members by {@code "pane"} while {@code fleet.leaders} has
+     * any entry, or any {@code fleet.collaborators} entry names a {@code tab}. A pane-placed
+     * member lands inside the focused tab rather than its own, so it can land inside a lead's or
+     * collaborator's own labelled tab. {@link dev.ltms.fleet.herdr.LeadTabScanner} identifies a
+     * lead or collaborator purely by that tab's label — it does not exclude the member space — so
+     * a member that ends up there, while its pane carries no entry in the spawned-member roster,
+     * is read back as that lead or collaborator and granted that identity's authority.
      *
-     * <p>Only a leader with a non-blank {@code tab} is in scope: one with no {@code tab} feeds
+     * <p>Every {@code fleet.leaders} entry is in scope regardless of its own {@code tab} field:
+     * {@link Leader#acceptedLabels()} always includes {@link Leader#LEAD_TAB_LABEL}. Only a
+     * collaborator with a non-blank {@code tab} is in scope: one with no {@code tab} feeds
      * nothing into {@link dev.ltms.fleet.herdr.LeadTabScanner}, so it creates no hazard here.
      *
-     * @throws IllegalStateException when any {@code profiles:} entry is pane-placed while any
-     *                               {@code fleet.leaders} entry names a non-blank {@code tab}
+     * @throws IllegalStateException when any {@code profiles:} entry is pane-placed while
+     *                               {@code fleet.leaders} is non-empty, or any
+     *                               {@code fleet.collaborators} entry names a non-blank
+     *                               {@code tab}
      */
     public void validatePanePlacementAgainstLeadTabs() {
-        if (fleet == null || fleet.leaders().isEmpty()) {
+        if (fleet == null) {
             return;
         }
-        boolean anyLeaderHasTab = fleet.leaders().values().stream()
-                .anyMatch(leader -> leader != null && leader.tab() != null && !leader.tab().isBlank());
-        if (!anyLeaderHasTab) {
+        boolean anyLead = !fleet.leaders().isEmpty();
+        boolean anyCollaboratorHasTab = fleet.collaborators().values().stream()
+                .anyMatch(c -> c != null && c.tab() != null && !c.tab().isBlank());
+        if (!anyLead && !anyCollaboratorHasTab) {
             return;
         }
         List<String> bad = new ArrayList<>();
@@ -2774,10 +2994,13 @@ public record FleetConfig(
             return;
         }
         throw new IllegalStateException("refusing to start: profile(s) " + bad
-                + " use placement: pane while fleet.leaders names a tab. A pane-placed member can "
-                + "land inside a lead's labelled tab and be read back as the lead, granted "
-                + "spawn/stop/send on the whole fleet. Set placement: tab for each named profile, "
-                + "or remove the tab from every fleet.leaders entry.");
+                + " use placement: pane while fleet.leaders or fleet.collaborators names a tab. A "
+                + "pane-placed member can land inside that labelled tab, and while its pane "
+                + "carries no entry in the spawned-member roster, it is read back as the lead or "
+                + "collaborator and granted that identity's authority. Set placement: tab for "
+                + "each named profile — the only fix when a lead triggered this, since a lead's "
+                + "tab label is fixed regardless of its own tab: field. A collaborator's tab can "
+                + "still be removed instead.");
     }
 
     /**
@@ -2800,15 +3023,6 @@ public record FleetConfig(
         }
     }
 
-    /** Case-insensitive prefix test that tolerates a null/blank label. */
-    private static boolean startsWithIgnoreCase(String label, String prefix) {
-        if (label == null || prefix == null || prefix.isBlank()) {
-            return false;
-        }
-        String stripped = label.strip();
-        return stripped.regionMatches(true, 0, prefix, 0, prefix.length());
-    }
-
     /**
      * Reject a subscription profile whose {@code env:} block tries to reseat the Anthropic binding
      * (CB-542).
@@ -2893,8 +3107,13 @@ public record FleetConfig(
      * so duplicates are unrepresentable by construction once loaded — and {@link #load(Path)}
      * already rejects a duplicated slot name at parse time, before the map collapses.
      *
-     * @throws IllegalStateException when a slot names no profile or an unknown one, or when a lead
-     *                               can be neither found nor created, naming the offending entry
+     * <p>A lead's {@code profile} is optional — a {@code profile}-less lead is still useful
+     * recognise-only. Also rejects a {@code fleet.collaborators} entry with no (or a blank)
+     * {@code tab}: a collaborator carries no other field at all, so a blank {@code tab} leaves
+     * nothing for the entry to mean.
+     *
+     * @throws IllegalStateException when a slot or a lead references an unknown profile, or a
+     *                               collaborator names no tab, naming the offending entry
      */
     public void validateMembers() {
         if (fleet == null) {
@@ -2927,9 +3146,15 @@ public record FleetConfig(
                         + "', which is not a configured profiles: entry (have: " + profiles.keySet()
                         + ").");
             }
-            if (leader.tab() == null || leader.tab().isBlank()) {
-                bad.add("fleet.leaders." + name + " has no tab: — a lead is now found (and, if "
-                        + "auto-launched, labelled) purely by its tab, so every entry must name one.");
+        });
+        fleet.collaborators().forEach((name, collaborator) -> {
+            if (collaborator == null) {
+                return;
+            }
+            if (collaborator.tab() == null || collaborator.tab().isBlank()) {
+                bad.add("fleet.collaborators." + name + " has no tab: — a collaborator is "
+                        + "recognised purely by its tab, and carries no other field, so every "
+                        + "entry must name one.");
             }
         });
         if (!bad.isEmpty()) {

fleetd/src/main/java/dev/ltms/fleet/herdr/HerdrRouter.java

@@ -11,12 +11,18 @@ public final class HerdrRouter implements AutoCloseable {
     private final AgentControl memberAgents;
     private final WorkspaceControl leadSpaces;
     private final WorkspaceControl memberSpaces;
-    private final Predicate<String> isLead;
+    private final Predicate<String> routeToLead;
 
-    public HerdrRouter(HerdrClient lead, HerdrClient member, Predicate<String> isLead) {
+    /**
+     * @param routeToLead true for a terminal whose pane lives in the lead herdr daemon — a lead's
+     *                     own pane or a configured collaborator's, both opened by a person at a
+     *                     terminal rather than spawned, so both are found in the lead daemon rather
+     *                     than the member one
+     */
+    public HerdrRouter(HerdrClient lead, HerdrClient member, Predicate<String> routeToLead) {
         this.lead = Objects.requireNonNull(lead, "lead");
         this.member = member != null ? member : lead;
-        this.isLead = Objects.requireNonNull(isLead, "isLead");
+        this.routeToLead = Objects.requireNonNull(routeToLead, "routeToLead");
         leadAgents = new AgentControl(this.lead);
         memberAgents = this.member == this.lead ? leadAgents : new AgentControl(this.member);
         leadSpaces = new WorkspaceControl(this.lead);
@@ -27,7 +33,7 @@ public final class HerdrRouter implements AutoCloseable {
     public WorkspaceControl leadSpaces() { return leadSpaces; }
     public AgentControl memberAgents() { return memberAgents; }
     public WorkspaceControl memberSpaces() { return memberSpaces; }
-    public AgentControl agentsFor(String targetId) { return isLead.test(targetId) ? leadAgents : memberAgents; }
+    public AgentControl agentsFor(String targetId) { return routeToLead.test(targetId) ? leadAgents : memberAgents; }
 
     HerdrClient leadClient() { return lead; }
     HerdrClient memberClient() { return member; }

fleetd/src/main/java/dev/ltms/fleet/herdr/LeadTabScanner.java

@@ -25,14 +25,12 @@ import java.util.function.Supplier;
  * by first starting the session and asking it. Scanning closes that loop: label the tab, and the
  * pane is recognised on the next resolve.
  *
- * <p><strong>CB-579 — matched by name, not prefix.</strong> This used to strip one shared
- * {@code tabPrefix} off a label to derive the lead's name, and merged a config-supplied
- * {@code terminal_id} pin over every scan result so the pin could never expire. Both are gone: each
- * lead now configures its own exact {@code tab} label ({@code fleet.leaders.<name>.tab}), so this
- * class is handed a {@code tab → name} map up front and matches labels against it exactly
- * (case-insensitively). There is no merge step — a scan result is the whole answer. That is the
- * fix for the bug this replaces: a {@code terminal_id} pin surviving in config after the pane it
- * named was gone, so the daemon kept treating a dead session as a live lead forever.
+ * <p><strong>Matched by label within a space, not by a shared prefix.</strong> Each lead's accepted
+ * labels (the fixed {@code lead} label, plus a deprecated {@code tab} when still configured) are
+ * matched exactly (case-insensitively) against tabs in that lead's own space only — a tab named
+ * {@code lead} in one space never resolves to another space's lead. A scan result is the whole
+ * answer; nothing is merged in from configuration between scans, so a tab that is gone drops out on
+ * the very next scan instead of lingering forever.
  *
  * <p><strong>Direction of trust.</strong> The label names the lead; it never <em>grants</em>
  * anything a pane could take for itself. Three properties keep that honest:
@@ -96,13 +94,20 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
 
     private static final Logger log = LoggerFactory.getLogger(LeadTabScanner.class);
 
+    /** What a matched tab names: a lead or a collaborator. */
+    private enum Kind { LEAD, COLLABORATOR }
+
+    /** One matched tab's name and what it names. */
+    private record Entry(String name, Kind kind) {}
+
     private final HerdrClient herdr;
-    private final Map<String, String> tabToName;
+    private final Map<String, Map<String, String>> leadLabelsBySpace;
+    private final Map<String, String> collaboratorTabToName;
     private final Set<String> excludedWorkspaceLabels;
     private final long ttlNanos;
     private final LongSupplier clock;
 
-    private Map<String, String> cached = Map.of();
+    private Map<String, Entry> cached = Map.of();
     private long scannedAtNanos;
     private boolean everScanned;
 
@@ -118,36 +123,82 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
     /**
      * @param herdr                   the herdr client to query ({@code workspace.list},
      *                                {@code tab.list}, {@code pane.list} — all read-only)
-     * @param tabToName               every configured lead's exact tab label → its name
-     *                                ({@code fleet.leaders.<name>.tab}), matched case-insensitively
+     * @param leadLabelsBySpace       each configured lead's accepted tab labels, keyed by the
+     *                                lead's own space label, then by label, to its name — matched
+     *                                case-insensitively on both the space and the label. A tab
+     *                                matches a lead only within that lead's own space
      * @param excludedWorkspaceLabels workspaces never scanned — the configured worker spaces
      * @param ttlNanos                how long a scan result is reused before the next one
      * @param clock                   nanosecond time source ({@code System::nanoTime} in production)
      */
-    public LeadTabScanner(HerdrClient herdr, Map<String, String> tabToName,
+    public LeadTabScanner(HerdrClient herdr, Map<String, Map<String, String>> leadLabelsBySpace,
+                          Set<String> excludedWorkspaceLabels, long ttlNanos, LongSupplier clock) {
+        this(herdr, leadLabelsBySpace, Map.of(), excludedWorkspaceLabels, ttlNanos, clock);
+    }
+
+    /**
+     * As {@link #LeadTabScanner(HerdrClient, Map, Set, long, LongSupplier)}, additionally scanning
+     * for configured collaborator tabs in the same pass.
+     *
+     * @param collaboratorTabToName every configured collaborator's exact tab label → its name
+     *                              ({@code fleet.collaborators.<name>.tab}), matched
+     *                              case-insensitively in any space
+     */
+    public LeadTabScanner(HerdrClient herdr, Map<String, Map<String, String>> leadLabelsBySpace,
+                          Map<String, String> collaboratorTabToName,
                           Set<String> excludedWorkspaceLabels, long ttlNanos, LongSupplier clock) {
         this.herdr = herdr;
-        this.tabToName = normalize(tabToName);
+        this.leadLabelsBySpace = buildLeadIndex(leadLabelsBySpace);
+        this.collaboratorTabToName = normalizedLabelMap(collaboratorTabToName);
         this.excludedWorkspaceLabels = excludedWorkspaceLabels == null
                 ? Set.of() : Set.copyOf(excludedWorkspaceLabels);
         this.ttlNanos = ttlNanos;
         this.clock = clock;
     }
 
-    /** Keys stripped and lower-cased once, so every lookup is a plain map hit. */
-    private static Map<String, String> normalize(Map<String, String> tabToName) {
-        if (tabToName == null || tabToName.isEmpty()) {
+    /**
+     * Space and label keys stripped and lower-cased once, so every lookup is a plain map hit. A
+     * space with no usable labels is simply absent — {@link #leadLabelsFor} then finds nothing for
+     * it, which is also what a space with a {@code null} label gets.
+     */
+    private static Map<String, Map<String, String>> buildLeadIndex(
+            Map<String, Map<String, String>> leadLabelsBySpace) {
+        Map<String, Map<String, String>> out = new LinkedHashMap<>();
+        if (leadLabelsBySpace == null) {
             return Map.of();
         }
-        Map<String, String> out = new LinkedHashMap<>();
-        tabToName.forEach((tab, name) -> {
-            if (tab != null && !tab.isBlank() && name != null && !name.isBlank()) {
-                out.put(tab.strip().toLowerCase(Locale.ROOT), name);
+        leadLabelsBySpace.forEach((space, labelsToName) -> {
+            if (space == null || space.isBlank()) {
+                return;
+            }
+            Map<String, String> normalized = normalizedLabelMap(labelsToName);
+            if (!normalized.isEmpty()) {
+                out.put(space.strip().toLowerCase(Locale.ROOT), normalized);
             }
         });
         return Collections.unmodifiableMap(out);
     }
 
+    private static Map<String, String> normalizedLabelMap(Map<String, String> labelToName) {
+        Map<String, String> out = new LinkedHashMap<>();
+        if (labelToName != null) {
+            labelToName.forEach((label, name) -> {
+                if (label != null && !label.isBlank() && name != null && !name.isBlank()) {
+                    out.put(label.strip().toLowerCase(Locale.ROOT), name);
+                }
+            });
+        }
+        return Collections.unmodifiableMap(out);
+    }
+
+    /** The accepted lead labels configured for {@code spaceLabel}, or an empty map for no match. */
+    private Map<String, String> leadLabelsFor(String spaceLabel) {
+        if (spaceLabel == null) {
+            return Map.of();
+        }
+        return leadLabelsBySpace.getOrDefault(spaceLabel.strip().toLowerCase(Locale.ROOT), Map.of());
+    }
+
     /**
      * The current {@code terminal_id → lead name} map, rescanning when the cache has expired.
      *
@@ -156,6 +207,29 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
      */
     @Override
     public synchronized Map<String, String> get() {
+        return byKind(refresh(), Kind.LEAD);
+    }
+
+    /**
+     * The current {@code terminal_id → collaborator name} map, sharing the same scan and cache as
+     * {@link #get()} — both kinds are matched in one pass, so this never costs a second herdr call.
+     */
+    public synchronized Map<String, String> collaborators() {
+        return byKind(refresh(), Kind.COLLABORATOR);
+    }
+
+    private static Map<String, String> byKind(Map<String, Entry> entries, Kind kind) {
+        Map<String, String> out = new LinkedHashMap<>();
+        entries.forEach((terminal, entry) -> {
+            if (entry.kind() == kind) {
+                out.put(terminal, entry.name());
+            }
+        });
+        return Collections.unmodifiableMap(out);
+    }
+
+    /** Rescans if the cache has expired, otherwise returns the cached answer. */
+    private Map<String, Entry> refresh() {
         long now = clock.getAsLong();
         if (everScanned && now - scannedAtNanos < ttlNanos) {
             return cached;
@@ -165,43 +239,45 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
         scannedAtNanos = now;
         everScanned = true;
         try {
-            Map<String, String> fresh = scan();
+            Map<String, Entry> fresh = scan();
             if (!fresh.equals(cached)) {
-                log.info("lead panes: {}", fresh);
+                log.info("lead/collaborator panes: {}", fresh);
             }
             cached = fresh;
         } catch (HerdrException e) {
-            log.warn("lead-tab scan failed, keeping the {} lead(s) already known: {}",
+            log.warn("lead-tab scan failed, keeping the {} entr(y/ies) already known: {}",
                     cached.size(), e.getMessage());
         }
         return cached;
     }
 
     /** One full pass: labelled tabs → live agents in them → those panes' terminals. */
-    private Map<String, String> scan() {
-        Map<String, String> nameByTab = new LinkedHashMap<>();
+    private Map<String, Entry> scan() {
+        Map<String, Entry> entryByTab = new LinkedHashMap<>();
         for (JsonNode w : herdr.call("workspace.list").path("workspaces")) {
             Workspace ws = Workspace.from(w);
             if (ws.workspaceId() == null || excludedWorkspaceLabels.contains(ws.label())) {
                 continue;
             }
+            Map<String, String> leadLabelsHere = leadLabelsFor(ws.label());
             for (JsonNode t : herdr.call("tab.list", Map.of("workspace_id", ws.workspaceId())).path("tabs")) {
                 Tab tab = Tab.from(t);
-                String name = leadNameOf(tab.label());
-                if (name != null && tab.tabId() != null) {
-                    nameByTab.put(tab.tabId(), name);
+                Entry entry = entryOf(tab.label(), leadLabelsHere);
+                if (entry != null && tab.tabId() != null) {
+                    entryByTab.put(tab.tabId(), entry);
                 }
             }
         }
 
-        if (nameByTab.isEmpty()) {
+        if (entryByTab.isEmpty()) {
             gracedTerminals = Set.of();
             return Map.of();
         }
 
-        // fleetd #359: a labelled tab is only a lead when herdr also reports a running agent in
-        // it — the same liveness signal LeadLauncher.countLeads trusts for the identical purpose.
-        // Without this, a tab left behind by a session that has since died reads as live forever.
+        // fleetd #359: a labelled tab is only a lead (or collaborator) when herdr also reports a
+        // running agent in it — the same liveness signal LeadLauncher.countLeads trusts for the
+        // identical purpose. Without this, a tab left behind by a session that has since died reads
+        // as live forever.
         Set<String> tabsWithAgent = new HashSet<>();
         for (JsonNode a : herdr.call("agent.list").path("agents")) {
             String tabId = a.path("tab_id").asText(null);
@@ -210,18 +286,18 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
             }
         }
 
-        Map<String, String> byTerminal = new LinkedHashMap<>();
+        Map<String, Entry> byTerminal = new LinkedHashMap<>();
         Set<String> stillGraced = new HashSet<>();
         // One pane.list for every tab: panes carry tab_id, so the join is local.
         for (JsonNode p : herdr.call("pane.list", Map.of()).path("panes")) {
             String tabId = p.path("tab_id").asText(null);
-            String name = nameByTab.get(tabId);
+            Entry entry = entryByTab.get(tabId);
             String terminal = p.path("terminal_id").asText(null);
-            if (name == null || terminal == null || terminal.isBlank()) {
+            if (entry == null || terminal == null || terminal.isBlank()) {
                 continue;
             }
             if (tabsWithAgent.contains(tabId)) {
-                byTerminal.put(terminal, name);
+                byTerminal.put(terminal, entry);
                 continue;
             }
             // No agent reported for this tab, but its tab/pane are still here — this is the
@@ -230,7 +306,7 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
             // reported as live; a terminal we never reported live gets none, so the original #359
             // fix (a genuinely dead tab is never reported) is unaffected for the common case.
             if (cached.containsKey(terminal) && !gracedTerminals.contains(terminal)) {
-                byTerminal.put(terminal, name);
+                byTerminal.put(terminal, entry);
                 stillGraced.add(terminal);
             }
         }
@@ -239,18 +315,27 @@ public final class LeadTabScanner implements Supplier<Map<String, String>> {
     }
 
     /**
-     * The lead name a tab label declares, or {@code null} if it names none of the configured leads.
+     * The entry a tab label declares within one space, or {@code null} if it names neither a lead
+     * accepted in {@code leadLabelsHere} nor a configured collaborator.
      *
-     * <p>Exact match (case-insensitive, ends stripped) against {@link #tabToName} — no prefix
-     * stripping, so an operator's {@code "lead: something-else"} tab is never mistaken for a
-     * configured lead just because it shares a prefix. The match strips a trailing
-     * {@link PendingCloseMarker} first, so a tab {@code LeadLauncher} has flagged as maybe-dead but
-     * not yet closed keeps resolving normally while that reconcile is pending.
+     * <p>Exact match (case-insensitive, ends stripped) — no prefix stripping, so an operator's
+     * {@code "lead: something-else"} tab is never mistaken for a configured lead just because it
+     * shares a prefix. The match strips a trailing {@link PendingCloseMarker} first, so a tab
+     * {@code LeadLauncher} has flagged as maybe-dead but not yet closed keeps resolving normally
+     * while that reconcile is pending. A lead match wins over a collaborator match for the same
+     * label — a lead can already do everything a collaborator can, and config validation refuses a
+     * lead and a collaborator sharing one exact tab in the first place.
      */
-    private String leadNameOf(String label) {
+    private Entry entryOf(String label, Map<String, String> leadLabelsHere) {
         if (label == null) {
             return null;
         }
-        return tabToName.get(PendingCloseMarker.strip(label).toLowerCase(Locale.ROOT));
+        String normalized = PendingCloseMarker.strip(label).toLowerCase(Locale.ROOT);
+        String leadName = leadLabelsHere.get(normalized);
+        if (leadName != null) {
+            return new Entry(leadName, Kind.LEAD);
+        }
+        String collaboratorName = collaboratorTabToName.get(normalized);
+        return collaboratorName == null ? null : new Entry(collaboratorName, Kind.COLLABORATOR);
     }
 }

fleetd/src/main/java/dev/ltms/fleet/herdr/PaneLocator.java

@@ -4,6 +4,7 @@ import com.fasterxml.jackson.databind.JsonNode;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import java.util.LinkedHashMap;
 import java.util.LinkedHashSet;
 import java.util.List;
 import java.util.Map;
@@ -145,6 +146,52 @@ public final class PaneLocator {
         return ancestry;
     }
 
+    /**
+     * Every tab herdr tracks across every searched daemon, keyed by tab id, to its display label —
+     * the pane-discovery surface behind {@code GET /agents} and {@code fleet_list}'s {@code panes}
+     * row. Collapses to one scan in the single-daemon deployment, the same as
+     * {@link #terminalForPid}. A tab herdr reports with no label maps to a {@code null} value here;
+     * a tab with no {@code tab_id} is skipped.
+     */
+    public Map<String, String> tabLabelsByTabId() {
+        Map<String, String> out = new LinkedHashMap<>();
+        for (HerdrClient herdr : herdrs) {
+            for (JsonNode w : herdr.call("workspace.list").path("workspaces")) {
+                String workspaceId = w.path("workspace_id").asText(null);
+                if (workspaceId == null) {
+                    continue;
+                }
+                for (JsonNode t : herdr.call("tab.list", Map.of("workspace_id", workspaceId)).path("tabs")) {
+                    Tab tab = Tab.from(t);
+                    if (tab.tabId() != null) {
+                        out.put(tab.tabId(), tab.label());
+                    }
+                }
+            }
+        }
+        return out;
+    }
+
+    /**
+     * Every workspace ("space") herdr tracks across every searched daemon, keyed by workspace id,
+     * to its display label — the human-readable name behind {@code fleet_list}'s {@code panes} row,
+     * next to herdr's own internal {@code workspaceId}. Collapses to one scan in the single-daemon
+     * deployment, the same as {@link #terminalForPid}. A workspace herdr reports with no label maps
+     * to a {@code null} value here; a workspace with no {@code workspace_id} is skipped.
+     */
+    public Map<String, String> workspaceLabelsByWorkspaceId() {
+        Map<String, String> out = new LinkedHashMap<>();
+        for (HerdrClient herdr : herdrs) {
+            for (JsonNode w : herdr.call("workspace.list").path("workspaces")) {
+                Workspace workspace = Workspace.from(w);
+                if (workspace.workspaceId() != null) {
+                    out.put(workspace.workspaceId(), workspace.label());
+                }
+            }
+        }
+        return out;
+    }
+
     /** Whether a pane owns one of the scanned pid's ancestors, or the check of it failed outright. */
     private enum Ownership { OWNS, DOES_NOT_OWN, UNKNOWN }
 

fleetd/src/main/java/dev/ltms/fleet/herdr/PromptBox.java

@@ -0,0 +1,169 @@
+package dev.ltms.fleet.herdr;
+
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.util.List;
+import java.util.Map;
+import java.util.concurrent.ConcurrentHashMap;
+
+/**
+ * Whether an agent pane's input box is clear for a delivery.
+ *
+ * <p>{@link AgentControl#send} pastes its text and submits it in the same call, so a delivery into a
+ * pane whose input box already holds characters submits those characters too. {@link
+ * AgentStatus#injectable()} cannot see that: it describes the agent, and an agent waiting at its
+ * prompt reports the same status whether its box is empty or holds a half-typed line. This reads the
+ * box itself.
+ *
+ * <p>Only a box that is positively empty clears the gate. A box with content, a pane this cannot
+ * recognise, and a failed read all hold the delivery, because a held delivery is recoverable and a
+ * submitted half-line is not. Every caller must therefore be a path that retries.
+ *
+ * <p>A pane that holds for {@link #HOLD_WARN_STREAK} consecutive checks gets one warning, so a box
+ * that never clears is visible instead of silent. The warning repeats only after the box has cleared
+ * again.
+ */
+public final class PromptBox {
+
+    private static final Logger log = LoggerFactory.getLogger(PromptBox.class);
+
+    /**
+     * herdr {@code agent.read} source. {@code detection} is the region herdr itself uses for status
+     * detection, so the input box is always drawn in it. It is not a short tail: it carries transcript
+     * scrollback above the box, including earlier prompts the operator has already submitted, which is
+     * why only the last box line on it is the live one.
+     */
+    static final String PROBE_SOURCE = "detection";
+
+    /** Consecutive holds for one target before one warning is logged. */
+    static final int HOLD_WARN_STREAK = 20;
+
+    /**
+     * Input box markers, each matched only as a line's first characters: the caret the current TUI
+     * draws, and the bordered box an older one drew. A marker further along a line is transcript text,
+     * such as a caret inside something the operator quoted.
+     */
+    private static final List<String> BOX_MARKERS = List.of("❯", "│ >");
+
+    /** Marker of a turn that is still generating; a box drawn under it is not a settled prompt. */
+    private static final String ACTIVE_TURN_MARKER = "esc to interrupt";
+
+    /** Block glyphs a terminal capture can leave in an otherwise empty box for the cursor cell. */
+    private static final String CURSOR_GLYPHS = "█▉▊▋▌▍▎▏";
+
+    /** What a box holds: nothing, unsubmitted characters, or a pane this cannot read as a box. */
+    public enum State { EMPTY, DRAFT, UNREADABLE }
+
+    /** A box reading: its state, and how many characters it holds ({@code 0} unless {@code DRAFT}). */
+    public record Reading(State state, int characters) {
+    }
+
+    private final AgentControl agents;
+
+    /** Consecutive holds per target, so a box that never clears can be warned about once. */
+    private final Map<String, Integer> holdStreaks = new ConcurrentHashMap<>();
+
+    public PromptBox(AgentControl agents) {
+        this.agents = agents;
+    }
+
+    /**
+     * Whether {@code target}'s input box is empty, so a delivery would submit only its own text.
+     * {@code false} means hold and come back; it never means the delivery failed.
+     */
+    public boolean clearToSubmit(String target) {
+        Reading reading = inspect(target);
+        if (reading.state() == State.EMPTY) {
+            holdStreaks.remove(target);
+            return true;
+        }
+        int streak = holdStreaks.merge(target, 1, Integer::sum);
+        if (streak == HOLD_WARN_STREAK) {
+            log.warn("prompt box of {} has held a delivery {} times in a row ({}, {} character(s) in the box)"
+                            + " — nothing is lost, delivery resumes once the box is empty",
+                    target, streak, reading.state(), reading.characters());
+        } else {
+            log.debug("prompt box of {} is {} ({} character(s)), holding delivery {}",
+                    target, reading.state(), reading.characters(), streak);
+        }
+        return false;
+    }
+
+    /** Read and classify {@code target}'s pane. A read failure reads as {@link State#UNREADABLE}. */
+    private Reading inspect(String target) {
+        String pane;
+        try {
+            pane = agents.read(target, PROBE_SOURCE);
+        } catch (RuntimeException e) {
+            log.debug("prompt box read for {} failed, holding delivery: {}", target, e.getMessage());
+            return new Reading(State.UNREADABLE, 0);
+        }
+        return classify(pane);
+    }
+
+    /**
+     * Classify a Claude Code TUI pane region. Pure, so it is unit-testable without herdr.
+     *
+     * <p>{@link State#EMPTY} needs two positive signals: the pane's last box line holds nothing after
+     * its marker, and nothing below that line says a turn is still generating. Everything else is
+     * {@link State#UNREADABLE} — a blank capture, or a region with no box line at all — so a pane this
+     * does not understand holds the delivery rather than guessing it is safe.
+     *
+     * <p>The generating marker is looked for only from the box line down. Above it is scrollback, where
+     * an earlier turn's marker survives; treating that as a live turn would make {@link State#EMPTY}
+     * unreachable and hold every delivery forever.
+     *
+     * <p>Whitespace, a trailing box border and a cursor block count as nothing. Any other character
+     * counts as the operator's unsubmitted text, including a placeholder hint a future TUI might draw
+     * there; that direction holds a delivery it could have sent, which {@link #HOLD_WARN_STREAK} makes
+     * visible.
+     */
+    static Reading classify(String pane) {
+        if (pane == null || pane.isBlank()) return new Reading(State.UNREADABLE, 0);
+        int box = lastBoxLineStart(pane);
+        if (box < 0) return new Reading(State.UNREADABLE, 0);
+        String fromBox = pane.substring(box);
+        if (fromBox.toLowerCase().contains(ACTIVE_TURN_MARKER)) return new Reading(State.UNREADABLE, 0);
+        String content = boxContent(firstLine(fromBox));
+        return content.isEmpty() ? new Reading(State.EMPTY, 0) : new Reading(State.DRAFT, content.length());
+    }
+
+    /** Offset of the last line starting with a box marker, or {@code -1} if the region has none. */
+    private static int lastBoxLineStart(String pane) {
+        int found = -1;
+        for (int start = 0; start <= pane.length(); ) {
+            int end = pane.indexOf('\n', start);
+            String line = pane.substring(start, end < 0 ? pane.length() : end);
+            if (markerLength(line) > 0) found = start;
+            if (end < 0) break;
+            start = end + 1;
+        }
+        return found;
+    }
+
+    /** Length of the box marker this line starts with, or {@code 0} if it starts with none. */
+    private static int markerLength(String line) {
+        for (String marker : BOX_MARKERS) {
+            if (line.startsWith(marker)) return marker.length();
+        }
+        return 0;
+    }
+
+    private static String firstLine(String text) {
+        int newline = text.indexOf('\n');
+        return newline < 0 ? text : text.substring(0, newline);
+    }
+
+    /** The text the box holds: its own line after the marker, stripped of border, padding and cursor. */
+    private static String boxContent(String boxLine) {
+        String line = boxLine.substring(markerLength(boxLine)).stripTrailing();
+        if (line.endsWith("│")) line = line.substring(0, line.length() - 1);
+        StringBuilder content = new StringBuilder();
+        for (char c : line.toCharArray()) {
+            if (Character.isWhitespace(c) || CURSOR_GLYPHS.indexOf(c) >= 0) continue;
+            content.append(c);
+        }
+        return content.toString();
+    }
+}

fleetd/src/main/java/dev/ltms/fleet/herdr/ResilientAgentLaunch.java

@@ -0,0 +1,146 @@
+package dev.ltms.fleet.herdr;
+
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.nio.charset.StandardCharsets;
+import java.util.List;
+import java.util.function.IntFunction;
+
+/**
+ * The three protections every {@code agent.start} caller needs against herdr's pane-typed launch
+ * surface (fleetd #220, #727): a byte-limit check on the assembled command line, a bounded retry
+ * on {@code agent_pane_busy} (the target pane's shell has not reached its prompt yet), and a
+ * bounded retry on {@code agent_name_taken} with a fresh name each attempt. One implementation —
+ * every caller of {@code agent.start}, lead or member, goes through this seam rather than carrying
+ * its own copy.
+ */
+public final class ResilientAgentLaunch {
+
+    private static final Logger log = LoggerFactory.getLogger(ResilientAgentLaunch.class);
+
+    private ResilientAgentLaunch() {
+    }
+
+    /**
+     * The pty line buffer herdr types a launch command into: BSD/macOS {@code MAX_CANON}. Not a
+     * fleetd choice and not configurable — see {@link #checkFits}.
+     */
+    public static final int PANE_COMMAND_BYTE_LIMIT = 1024;
+
+    /** Per-argument allowance for the separating space and a shell quote pair fleetd cannot see. */
+    private static final int QUOTING_OVERHEAD_PER_ARG = 3;
+
+    /** herdr rejects a duplicate agent {@code name}; a caller retries a bumped name this many times. */
+    public static final int NAME_RETRIES = 8;
+
+    /**
+     * Retries for {@code agent.start} against a seed pane whose shell has not reached its prompt
+     * yet — {@code tab.create}/{@code pane.split} return as soon as the pane exists, and herdr
+     * refuses to start an agent in a pane that is not "an available shell" ({@code agent_pane_busy}).
+     */
+    public static final int SHELL_READY_RETRIES = 20;
+
+    /** Raised by {@link #checkFits} when the assembled command cannot fit the pane line. */
+    public static final class TooLargeException extends RuntimeException {
+        public TooLargeException(String message) {
+            super(message);
+        }
+    }
+
+    /**
+     * Verify the assembled launch line fits the pane herdr types it into. herdr does not exec the
+     * launch command — it TYPES it into the pane as one line, and a pty line buffer holds only
+     * {@value #PANE_COMMAND_BYTE_LIMIT} bytes. Everything past that byte is dropped with no error
+     * anywhere: herdr answers "agent started", the backend exits on the mangled argument it was
+     * handed, the pane closes, and the only symptom is a readiness timeout with no reason. That is
+     * how fleetd #214 broke every claude-code spawn — one 50-byte flag pushed a 978-byte command to
+     * 1028, and the tail that got cut was {@code --autocompact 250000}.
+     *
+     * <p>So measure it here and refuse, loudly and immediately, rather than start something that
+     * cannot work. The estimate is deliberately conservative: fleetd cannot see herdr's quoting, so
+     * every argument is charged its own bytes plus a separator and a quote pair. An over-estimate
+     * costs a clear error at a length that was already unsafe; an under-estimate would let the
+     * silent truncation back in.
+     *
+     * @param label names the launch in the refusal message (a profile name)
+     * @param argv  the full argv, including the executable at index 0
+     * @throws TooLargeException naming the limit, the estimate, and the longest argument
+     */
+    public static void checkFits(String label, List<String> argv) {
+        int bytes = 0;
+        String longest = null;
+        int longestBytes = 0;
+        for (String arg : argv) {
+            int argBytes = arg == null ? 0 : arg.getBytes(StandardCharsets.UTF_8).length;
+            bytes += argBytes + QUOTING_OVERHEAD_PER_ARG;
+            if (argBytes > longestBytes) {
+                longestBytes = argBytes;
+                longest = arg;
+            }
+        }
+        if (bytes <= PANE_COMMAND_BYTE_LIMIT) {
+            return;
+        }
+        String culprit = longest == null ? "<none>"
+                : longest.substring(0, Math.min(longest.length(), 60)) + (longest.length() > 60 ? "…" : "");
+        throw new TooLargeException(
+                "launch command for " + label + " is about " + bytes + " bytes, over the "
+                        + PANE_COMMAND_BYTE_LIMIT + "-byte limit of the pane line herdr types it into. "
+                        + "The pty would drop the tail silently and the backend would exit on a mangled "
+                        + "argument. Longest argument is " + longestBytes + " bytes: " + culprit
+                        + " — move it off the command line (a file flag) or shorten it.");
+    }
+
+    /**
+     * Start an agent into {@code paneId}, retrying {@code agent_pane_busy} up to {@code retries}
+     * times with {@code sleeper} run between attempts.
+     *
+     * @throws HerdrException the last {@code agent_pane_busy} failure once {@code retries} is
+     *                         spent, or immediately for any other herdr failure
+     */
+    public static Agent startAwaitingShellPrompt(AgentControl agents, String name, String kind,
+                                                  List<String> args, String paneId,
+                                                  int retries, Runnable sleeper) {
+        HerdrException busy = null;
+        for (int attempt = 0; attempt < retries; attempt++) {
+            try {
+                return agents.start(name, kind, args, paneId);
+            } catch (HerdrException e) {
+                if (!"agent_pane_busy".equals(e.code())) throw e;
+                log.debug("pane {} not at its shell prompt yet, retrying agent.start", paneId);
+                busy = e;
+                sleeper.run();
+            }
+        }
+        throw busy;
+    }
+
+    /**
+     * Start an agent under a freshly generated name each attempt, retrying {@code agent_name_taken}
+     * up to {@code nameRetries} times — herdr refuses a duplicate {@code name} outright, so a stale
+     * registry entry (a crashed session, a name the registry has not yet released) must not block a
+     * legitimate relaunch. Each attempt also carries its own {@link #startAwaitingShellPrompt} retry.
+     *
+     * @param nameForAttempt called once per attempt (0-based) to produce that attempt's name
+     * @throws HerdrException the last {@code agent_name_taken} failure once {@code nameRetries} is
+     *                         spent, or immediately for any other herdr failure
+     */
+    public static Agent startUniquelyNamed(AgentControl agents, String kind, List<String> args,
+                                            String paneId, IntFunction<String> nameForAttempt,
+                                            int nameRetries, int shellReadyRetries, Runnable sleeper) {
+        HerdrException last = null;
+        for (int attempt = 0; attempt < nameRetries; attempt++) {
+            String name = nameForAttempt.apply(attempt);
+            try {
+                return startAwaitingShellPrompt(agents, name, kind, args, paneId,
+                        shellReadyRetries, sleeper);
+            } catch (HerdrException e) {
+                if (!"agent_name_taken".equals(e.code())) throw e;
+                log.debug("agent name '{}' taken, retrying", name);
+                last = e;
+            }
+        }
+        throw last;
+    }
+}

fleetd/src/main/java/dev/ltms/fleet/inject/MemberPresence.java

@@ -4,16 +4,17 @@ import java.util.concurrent.ConcurrentHashMap;
 import java.util.Set;
 
 /**
- * Tracks which workers are <em>available</em> — their Claude has booted and connected its MCP client
- * to the bridge (CB-113). This is the reliable readiness signal, unlike herdr's {@code agent_status},
- * which reports {@code idle} for a worker whose Claude is still booting. Delivering into that boot
- * window pastes into a not-yet-ready TUI (the text is lost) and wedges the worker's delivery state,
- * so the {@link Injector} holds the first delivery until the worker is present here.
+ * Tracks which peers are <em>available</em> — their Claude has booted and connected its MCP
+ * client to the bridge. For a spawned member this is the reliable readiness signal,
+ * unlike herdr's {@code agent_status}, which reports {@code idle} while its Claude is still
+ * booting. Delivering into that boot window pastes into a not-yet-ready TUI (the text is lost)
+ * and wedges that member's delivery state, so the {@link Injector} holds a spawned member's
+ * first delivery until it is present here.
  *
- * <p>Populated from the MCP transport: any MCP request whose connection resolves to a worker terminal
- * marks that worker present (its {@code initialize} is the first such contact). A worker that never
- * mounts the bridge MCP is never marked present — its sends stay queued until they time out, which is
- * correct (it could not have replied anyway).
+ * <p>Populated from the MCP transport, for the peers whose deliverability rests on proving a live
+ * MCP contact rather than on a configured registry entry. A peer that never mounts the bridge MCP
+ * is never marked present — its sends stay queued until they time out, which is correct (it could
+ * not have replied anyway).
  */
 public class MemberPresence {
 

fleetd/src/main/java/dev/ltms/fleet/lead/LeadLauncher.java

@@ -5,6 +5,7 @@ import dev.ltms.fleet.herdr.Agent;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.HerdrException;
 import dev.ltms.fleet.herdr.PendingCloseMarker;
+import dev.ltms.fleet.herdr.ResilientAgentLaunch;
 import dev.ltms.fleet.herdr.Tab;
 import dev.ltms.fleet.herdr.Workspace;
 import dev.ltms.fleet.herdr.WorkspaceControl;
@@ -12,13 +13,16 @@ import dev.ltms.fleet.launch.ClaudeCodeArguments;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import java.security.SecureRandom;
 import java.util.ArrayList;
 import java.util.LinkedHashMap;
 import java.util.LinkedHashSet;
 import java.util.List;
+import java.util.Locale;
 import java.util.Map;
 import java.util.Objects;
 import java.util.Set;
+import java.util.concurrent.atomic.AtomicLong;
 import dev.ltms.fleet.peer.PeerLauncher;
 
 /**
@@ -80,9 +84,19 @@ public final class LeadLauncher {
 
     private static final Logger log = LoggerFactory.getLogger(LeadLauncher.class);
 
+    /** Attempts {@link #relaunch(String)} makes before giving up and returning {@code null}. */
+    static final int RELAUNCH_ATTEMPTS = 3;
+
     private final AgentControl agents;
     private final WorkspaceControl spaces;
     private final FleetConfig cfg;
+    private final Runnable sleeper;
+
+    // Per-process token mixed into each lead agent name so a fresh daemon process (seq back at 0)
+    // cannot collide with a same-name lead that outlived a restart — the same scheme
+    // HerdrPeerLauncher uses for members (fleetd #727).
+    private final String nameNonce = String.format("%06x", new SecureRandom().nextInt(1 << 24));
+    private final AtomicLong nameSeq = new AtomicLong();
 
     /**
      * @param agents herdr agent control (start, list)
@@ -90,9 +104,27 @@ public final class LeadLauncher {
      * @param cfg    the loaded config — {@code fleet.leaders}, {@code profiles} and each lead's tab
      */
     public LeadLauncher(AgentControl agents, WorkspaceControl spaces, FleetConfig cfg) {
+        this(agents, spaces, cfg, () -> sleepUninterruptibly(300));
+    }
+
+    /**
+     * Test seam: as above, plus an injectable {@code sleeper} for the {@code agent_pane_busy}
+     * retry (fleetd #727), so a test can prove the retry budget without a real sleep.
+     */
+    LeadLauncher(AgentControl agents, WorkspaceControl spaces, FleetConfig cfg, Runnable sleeper) {
         this.agents = agents;
         this.spaces = spaces;
         this.cfg = cfg;
+        this.sleeper = sleeper;
+    }
+
+    /** Uninterruptible sleep — the production {@link #sleeper} between {@code agent_pane_busy} retries. */
+    private static void sleepUninterruptibly(long ms) {
+        try {
+            Thread.sleep(ms);
+        } catch (InterruptedException e) {
+            Thread.currentThread().interrupt();
+        }
     }
 
     /**
@@ -173,24 +205,14 @@ public final class LeadLauncher {
                 log.info("lead '{}': {} live, {} wanted — nothing to start", name, running, wanted);
                 continue;
             }
-            if (!lead.isCreatable()) {
-                // A lead with a `tab:` but no `profile:` is recognise-only by design: the operator
-                // opens it by hand. Say so once rather than looking like a silent failure.
-                log.info("lead '{}' is not live, and names no profile — it can be recognised but not "
-                        + "launched. Add `profile:` under fleet.leaders.{} to have fleetd start it.",
-                        name, name);
-                continue;
-            }
 
-            FleetConfig.Profile profile = cfg.profiles().get(lead.profile());
-            if (profile == null) {
-                log.warn("lead '{}' names profile '{}', which is not configured — not launching",
-                        name, lead.profile());
+            ResolvedLead resolved = resolveLaunchable(name);
+            if (resolved == null) {
                 continue;
             }
 
             for (int i = running; i < wanted; i++) {
-                if (launch(name, lead, profile)) {
+                if (launch(name, resolved.lead(), resolved.profile()) != null) {
                     started++;
                 }
             }
@@ -198,17 +220,89 @@ public final class LeadLauncher {
         return started;
     }
 
+    /** A declared lead paired with the profile it launches on — {@link #resolveLaunchable}'s result. */
+    private record ResolvedLead(FleetConfig.Leader lead, FleetConfig.Profile profile) {
+    }
+
+    /**
+     * The declared {@code Leader} and its {@code Profile} for {@code name}, read from the config
+     * snapshot this launcher was constructed with.
+     *
+     * @return the resolved pair, or {@code null} (having logged) if {@code name} is not declared
+     *         under {@code fleet.leaders}, that lead names no {@code profile:} (a {@code tab:}-only,
+     *         recognise-only lead), or its {@code profile:} is not configured. Shared by
+     *         {@link #ensureLeads()} and {@link #relaunch(String)} so the three refusals and their
+     *         wording live in one place.
+     */
+    private ResolvedLead resolveLaunchable(String name) {
+        FleetConfig.Leader lead = cfg.fleet().leaders().get(name);
+        if (lead == null) {
+            log.warn("lead '{}' is not declared under fleet.leaders — not launching", name);
+            return null;
+        }
+        if (!lead.isCreatable()) {
+            // A lead with a `tab:` but no `profile:` is recognise-only by design: the operator
+            // opens it by hand. Say so once rather than looking like a silent failure.
+            log.info("lead '{}' names no profile — it can be recognised but not launched. Add "
+                    + "`profile:` under fleet.leaders.{} to have fleetd start it.", name, name);
+            return null;
+        }
+
+        FleetConfig.Profile profile = cfg.profiles().get(lead.profile());
+        if (profile == null) {
+            log.warn("lead '{}' names profile '{}', which is not configured — not launching",
+                    name, lead.profile());
+            return null;
+        }
+        return new ResolvedLead(lead, profile);
+    }
+
+    /**
+     * Start the named lead from the config snapshot this launcher was constructed with — not a
+     * live read, so a lead's {@code profile:} or {@code tab:} edited in config needs a daemon
+     * restart to take effect here — outside of {@link #ensureLeads()}'s {@code instances}
+     * bookkeeping.
+     *
+     * @return the started {@link Agent}, or {@code null} if {@code name} is not declared under
+     *         {@code fleet.leaders}, that lead names no {@code profile:} (a {@code tab:}-only,
+     *         recognise-only lead), its {@code profile:} is not configured, or every attempt up to
+     *         {@link #RELAUNCH_ATTEMPTS} failed to start it. Never throws.
+     *
+     * <p>Does not count how many instances of this lead are already live. {@link #ensureLeads()}'s
+     * count exists to avoid starting a second orchestrator; the caller of this method has already
+     * decided to replace the lead and owns that decision.
+     *
+     * <p>Retries the whole launch attempt — not only the {@code agent_name_taken}/
+     * {@code agent_pane_busy} cases {@link ResilientAgentLaunch} already retries inside one
+     * {@code agents.start} call — up to {@link #RELAUNCH_ATTEMPTS} times, sleeping via the
+     * injected sleeper between attempts, and returns the agent from the first attempt that
+     * succeeds.
+     */
+    public Agent relaunch(String name) {
+        ResolvedLead resolved = resolveLaunchable(name);
+        if (resolved == null) {
+            return null;
+        }
+
+        for (int attempt = 1; attempt <= RELAUNCH_ATTEMPTS; attempt++) {
+            Agent started = launch(name, resolved.lead(), resolved.profile());
+            if (started != null) {
+                return started;
+            }
+            if (attempt < RELAUNCH_ATTEMPTS) {
+                sleeper.run();
+            }
+        }
+        return null;
+    }
+
     /**
      * How many live leads exist per configured name, and which of that name's labelled tabs are
-     * <em>not</em> live: a running agent in a tab labelled with that lead's exact {@code tab}
-     * (CB-579). A member sitting in the same shared workspace is not counted as a lead because its
-     * tab carries a different label, not because any workspace is excluded from this count.
-     *
-     * <p>There used to be a second path here — a running agent on the terminal a
-     * {@code fleet.leaders.<name>.terminal} pin named, for a lead opened and pinned by hand. That
-     * pin is retired: {@code tab} is now the only field identity depends on, and {@link Agent}
-     * already carries {@link Agent#tabId()} directly, so a hand-opened lead is found the same way an
-     * auto-launched one is — by labelling its tab to match.
+     * <em>not</em> live: a running agent in a tab, in that lead's own space, carrying one of its
+     * {@link FleetConfig.Leader#acceptedLabels()}. A member sitting in the same shared workspace is
+     * not counted as a lead because its tab carries a different label, not because any workspace is
+     * excluded from this count. A tab matching a lead's label in a <em>different</em> space is not
+     * counted either — space is the uniqueness boundary between leads.
      *
      * <p>fleetd #359 review finding 1: a labelled tab with nothing running in it is split into
      * {@code toClose} (already flagged pending-close by a previous reconcile, and still dead — two
@@ -222,12 +316,11 @@ public final class LeadLauncher {
     }
 
     private Map<String, LeadCount> countLeads(Map<String, FleetConfig.Leader> leaders) {
-        // A lead and the members share ONE workspace now (the operator asked for a single "session"
-        // with many tabs), so a workspace can no longer be excluded wholesale — the lead lives in the
-        // member workspace by design. The sole discriminator is the exact tab label: a lead carries
-        // its configured `fleet.leaders.<name>.tab` ("lead: opus"), while a member carries its
-        // profile's `worker: {profile} #{n}` template. These never collide, so an exact-label match
-        // separates them without needing to know which workspace anyone is in.
+        // A lead and the members share ONE workspace (the operator asked for a single "session" with
+        // many tabs), so a workspace can no longer be excluded wholesale — the lead lives in the
+        // member workspace by design. The discriminator is the tab label together with the space: a
+        // member's tab never carries one of a lead's accepted labels, and a lead's own label only
+        // counts within that lead's configured space.
         Map<String, String> nameByTab = new LinkedHashMap<>();
         Set<String> flaggedTabIds = new LinkedHashSet<>();
         for (Workspace ws : spaces.listWorkspaces()) {
@@ -235,7 +328,7 @@ public final class LeadLauncher {
                 continue;
             }
             for (Tab tab : spaces.listTabs(ws.workspaceId())) {
-                String declared = leadNameOf(tab.label(), leaders);
+                String declared = leadNameOf(tab.label(), ws.label(), leaders);
                 if (declared != null && tab.tabId() != null) {
                     nameByTab.put(tab.tabId(), declared);
                     if (PendingCloseMarker.isFlagged(tab.label())) {
@@ -285,29 +378,41 @@ public final class LeadLauncher {
     }
 
     /**
-     * The configured lead a tab label names, or {@code null} for a label that names none.
+     * The configured lead a tab names, or {@code null} for a label or space that names none.
      *
-     * <p>Matched exactly (case-insensitively) against each lead's configured {@code tab}, so an
-     * operator's {@code "lead: something-else"} tab is not mistaken for a configured lead. A
-     * trailing {@link PendingCloseMarker} is stripped first, so a tab this class flagged on a
-     * previous reconcile is still recognised as the same lead's tab on this one.
+     * <p>A match requires both: the label (case-insensitively, trailing {@link PendingCloseMarker}
+     * stripped) must be one of the lead's {@link FleetConfig.Leader#acceptedLabels()}, and {@code
+     * space} must be that lead's own {@link FleetConfig.Leader#workspace()}. The same label in a
+     * different space names no lead — space is the uniqueness boundary between leads.
      */
-    private String leadNameOf(String label, Map<String, FleetConfig.Leader> leaders) {
-        if (label == null) {
+    private String leadNameOf(String label, String space, Map<String, FleetConfig.Leader> leaders) {
+        if (label == null || space == null) {
             return null;
         }
-        String l = PendingCloseMarker.strip(label);
+        String l = PendingCloseMarker.strip(label).toLowerCase(Locale.ROOT);
         for (Map.Entry<String, FleetConfig.Leader> e : leaders.entrySet()) {
-            String tab = e.getValue().tabLabel();
-            if (tab != null && l.equalsIgnoreCase(tab.strip())) {
+            FleetConfig.Leader lead = e.getValue();
+            if (lead == null || !lead.workspace().equalsIgnoreCase(space)) {
+                continue;
+            }
+            if (lead.acceptedLabels().contains(l)) {
                 return e.getKey();
             }
         }
         return null;
     }
 
-    /** Start one lead. Returns false (having logged) rather than throwing on any failure. */
-    private boolean launch(String name, FleetConfig.Leader lead, FleetConfig.Profile profile) {
+    /**
+     * Start one lead. Returns null (having logged) rather than throwing on any failure.
+     *
+     * <p>Goes through the same {@link ResilientAgentLaunch} seam every member spawn uses
+     * (fleetd #727): the assembled argv is refused outright if it cannot fit the pane line herdr
+     * types it into, a stale {@code agent_name_taken} (a crashed session's name the registry has
+     * not yet released) is retried under a fresh per-attempt name rather than refusing the whole
+     * relaunch, and a seed pane whose shell has not reached its prompt yet ({@code
+     * agent_pane_busy}) is retried rather than failing on the first miss.
+     */
+    private Agent launch(String name, FleetConfig.Leader lead, FleetConfig.Profile profile) {
         String label = lead.tabLabel();
         String cwd = (lead.cwd() == null || lead.cwd().isBlank())
                 ? System.getProperty("user.dir") : lead.cwd();
@@ -323,8 +428,12 @@ public final class LeadLauncher {
             // Same shape as the member launchers: herdr resolves the executable from `kind`, so
             // argv[0] (the configured launcher, e.g. `ccs`) is dropped and only the rest is passed.
             List<String> argv = leadArgv(profile);
-            Agent started = agents.start("lead-" + name, herdrKind(profile),
-                    argv.isEmpty() ? argv : argv.subList(1, argv.size()), tab.rootPaneId());
+            ResilientAgentLaunch.checkFits(profile.profile(), argv);
+            List<String> args = argv.isEmpty() ? argv : argv.subList(1, argv.size());
+            Agent started = ResilientAgentLaunch.startUniquelyNamed(agents, herdrKind(profile), args,
+                    tab.rootPaneId(),
+                    attempt -> "lead-" + name + "-" + nameNonce + "-" + nameSeq.incrementAndGet(),
+                    ResilientAgentLaunch.NAME_RETRIES, ResilientAgentLaunch.SHELL_READY_RETRIES, sleeper);
 
             // Label AFTER the start succeeds. A label written before would survive a failed start
             // and then read back as a live lead on the next boot, which is the exact staleness the
@@ -334,7 +443,7 @@ public final class LeadLauncher {
             log.info("lead '{}' launched: profile={} tab={} pane={} terminal={} label='{}' cwd={}",
                     name, profile.profile(), tab.tab().tabId(), started.paneId(),
                     started.terminalId(), label, cwd);
-            return true;
+            return started;
         } catch (RuntimeException e) {
             log.warn("lead '{}' failed to launch on profile '{}': {}",
                     name, profile.profile(), e.getMessage());
@@ -346,7 +455,7 @@ public final class LeadLauncher {
                             tab.tab().tabId(), cleanup.getMessage());
                 }
             }
-            return false;
+            return null;
         }
     }
 

fleetd/src/main/java/dev/ltms/fleet/lead/LeadRollover.java

@@ -1,8 +1,11 @@
 package dev.ltms.fleet.lead;
 
 import dev.ltms.fleet.config.FleetConfig;
+import dev.ltms.fleet.herdr.Agent;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
+import dev.ltms.fleet.herdr.HerdrException;
+import dev.ltms.fleet.herdr.WorkspaceControl;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
@@ -24,64 +27,73 @@ import java.util.function.Supplier;
 /**
  * fleetd #480: replace a lead session that has decided it is ready to be rolled over, without an
  * operator doing it by hand. A lead writes a handover file, calls {@link #open}, and then — once
- * every gate ({@link #confirm}'s own checks) has passed — a deferred, single-shot continuation
- * clears the lead's own pane and bootstraps a fresh session against that file.
+ * every gate ({@link #confirm}'s own checks) has passed — a deferred, single-shot continuation ends
+ * the lead's own pane, launches a fresh one, and bootstraps that fresh session against the file.
  *
  * <p>This is the executor behind the {@code fleet_handover} MCP tool ({@code
  * dev.ltms.fleet.mcp.FleetMcp#handover}), which drives {@link #open}, {@link #confirm}, {@link
- * #cancel}, and {@link #status} from a tool call — wired in fleetd #480 Unit C. <strong>An earlier
- * version of this paragraph said nothing called this class at all; that stopped being true once
- * that unit landed, and this correction exists so the javadoc does not go on claiming it.</strong>
+ * #cancel}, and {@link #status} from a tool call.
  *
- * <p><strong>{@code confirm()} cannot roll inline — a fleetd #480 correction.</strong> The first
- * version of this class called {@code agents.send(lead, "/clear")} directly from inside {@code
- * confirm()}, then polled for the pane to become injectable again. That is wrong, because {@code
- * confirm()} is called BY the lead, FROM the lead's own turn: the lead's pane is {@code WORKING}
- * for the whole duration of that call and cannot possibly report injectable until {@code confirm()}
- * itself returns. The poll always timed out — but only after the {@code /clear} had already been
- * sent and queued in the pane, where it fired the instant the turn ended anyway. The result was the
- * worst outcome this feature can produce: a silently destroyed lead context with no fresh session
- * ever started, and a refusal return value that claimed nothing had happened.
- *
- * <p>The fix: {@link #confirm} validates every gate, then does no I/O against the lead's own pane
- * at all — it only records that the request is approved and hands a one-shot continuation to
- * {@code continuationRunner} before returning. That continuation is what actually touches the pane,
- * once the calling turn has ended, in this order:
+ * <p><strong>{@code confirm()} cannot roll inline.</strong> {@code confirm()} is called BY the
+ * lead, FROM the lead's own turn: the lead's pane is {@code WORKING} for the whole duration of that
+ * call and cannot possibly report a real turn boundary until {@code confirm()} itself returns. So
+ * {@link #confirm} validates every gate, then does no I/O against the lead's own pane at all — it
+ * only records that the request is approved and hands a one-shot continuation to {@code
+ * continuationRunner} before returning. That continuation is what actually touches the pane, once
+ * the calling turn has ended, in this order:
  * <ol>
  *   <li>wait for the lead's own pane to report a real turn boundary — {@code IDLE} or {@code
  *       DONE}, never merely {@code BLOCKED} — i.e. wait for the very {@code confirm()} call that
  *       approved this roll to finish its turn — bounded by {@code turnSettleSeconds}. <strong>If
- *       this never happens, nothing else in this list runs: no {@code /clear} is ever sent.</strong>
- *       A lead that never goes idle is a lead still doing real work, and clearing it would throw
- *       away live context — exactly the failure this correction exists to prevent.</li>
- *   <li>{@code agents.send(lead, "/clear")}</li>
- *   <li>wait for {@code /clear} to be picked up and settle, bounded by {@code clearSettleSeconds}
- *       (fleetd #489: no longer a plain re-check of the same boundary — {@code /clear} starts no
- *       turn of its own, so this instead nudges the submit keystroke while no pickup has been seen,
- *       then waits for a real {@code WORKING} → {@code IDLE}/{@code DONE} boundary once one has;
- *       see {@link #waitForClearPickupAndSettle})</li>
- *   <li>{@code agents.send(lead, cfg.bootstrapTextFor(p.handoverPath()))}</li>
+ *       this never happens, nothing else in this list runs: the old pane is never touched.</strong>
+ *       A lead that never goes idle is a lead still doing real work, and tearing it down would throw
+ *       away live context.</li>
+ *   <li>capture the old pane id (and, through it, the old tab) from {@link AgentControl#get}, with
+ *       a bounded retry — the terminal-to-pane lookup it goes through can itself report a genuinely
+ *       live agent as not found (see {@code AgentControl#agentCall}'s own re-resolve-once
+ *       behaviour), and one false negative here must not abort an otherwise-healthy roll. Neither id
+ *       is ever re-resolved from the terminal again after this — once the pane below is closed there
+ *       is nothing left to resolve it from.</li>
+ *   <li>resolve the lead's configured name from its terminal, for the relaunch step below.</li>
+ *   <li>end the old session: close the pane (an already-gone pane counts as success; any other
+ *       failure propagates), then close its tab only when the pane was that tab's sole occupant —
+ *       the same pane-then-tab teardown {@code HerdrPeerLauncher#stop} uses for a member.</li>
+ *   <li>confirm the old pane is actually gone by polling {@link
+ *       dev.ltms.fleet.herdr.WorkspaceControl#locatePane} for a {@code null} result — never {@link
+ *       AgentControl#status}, and never the live-lead terminal map, each of which answers a
+ *       different question. <strong>If the old pane is never confirmed gone, no relaunch is
+ *       attempted</strong> — see {@link RollState#OLD_PANE_NEVER_DIED}.</li>
+ *   <li>launch a fresh lead with {@code LeadLauncher#relaunch}. <strong>If every attempt fails,
+ *       {@code bootstrapText} is never sent</strong> — see {@link RollState#RELAUNCH_FAILED}.</li>
+ *   <li>wait for the fresh pane to reach a real turn boundary ({@code IDLE} or {@code DONE},
+ *       never merely {@code BLOCKED}), bounded by {@code relaunchReadySeconds}. This is the
+ *       safety gate: typing into a pane that has not actually finished booting loses the
+ *       keystrokes. <strong>If the pane never becomes ready, {@code bootstrapText} is never
+ *       sent</strong> — see {@link RollState#RELAUNCH_NEVER_READY}.</li>
+ *   <li>wait for the fresh terminal to be recognised as a live lead — present in the live-lead
+ *       terminal map — bounded by {@code relaunchReadySeconds}. This is bookkeeping, not a
+ *       safety gate: {@code bootstrapText} is sent either way once the pane is ready, whether or
+ *       not this wait itself times out — see {@link RollState#RELAUNCH_NOT_RECOGNISED}.</li>
+ *   <li>{@code agents.send(newTerminal, cfg.bootstrapTextFor(p.handoverPath()))} — sent to the
+ *       FRESH terminal, never the one that was just torn down.</li>
  * </ol>
  * A {@link #confirm} that returns {@link RollDecision#approved()} therefore means <em>"every gate
- * passed and the roll is scheduled"</em>, never <em>"the pane has been cleared"</em> — the pane may
- * still be mid-turn, possibly for a long time, when the caller gets that answer back.
+ * passed and the roll is scheduled"</em>, never <em>"the lead has already been replaced"</em> — the
+ * old pane may still be mid-turn, possibly for a long time, when the caller gets that answer back.
  *
- * <p><strong>The safety invariant survives this change, restated precisely.</strong> The ticket
- * that first defined this class required "no timer, no scheduler, no background thread" so that
- * nothing but an explicit {@link #confirm} call could ever cause a {@code /clear}. That invariant
- * is about INITIATIVE, not about synchronicity, and this correction keeps it: {@code
- * continuationRunner} launches a single-shot task that exists only because one specific,
+ * <p><strong>The safety invariant.</strong> "No timer, no scheduler, no background thread" means
+ * that nothing but an explicit {@link #confirm} call can ever tear a lead's pane down.
+ * {@code continuationRunner} launches a single-shot task that exists only because one specific,
  * already-approved {@link #confirm} call created it — it is not recurring, it is not started at
  * construction time or on any schedule, and no two invocations of it ever share state. A recurring
- * heartbeat or timer that could decide on its own initiative to roll a pane is still, and will
- * always be, absent from this class. <strong>Nothing but an explicit {@link #confirm} call that
- * passes every gate can ever cause a {@code /clear} — that call may simply finish its own work
- * slightly later than the method return, as a continuation of the same approved request, rather
- * than entirely inside the method body.</strong>
+ * heartbeat or timer that could decide on its own initiative to roll a pane is absent from this
+ * class. <strong>Nothing but an explicit {@link #confirm} call that passes every gate can ever tear
+ * a pane down — that call may simply finish its own work slightly later than the method return, as
+ * a continuation of the same approved request, rather than entirely inside the method body.</strong>
  *
  * <p><strong>Identity is resolved by the caller, never looked up here — a second fleetd #480
  * correction.</strong> The first version resolved the pane to clear via {@code
- * PrimaryRegistry#primaryTerminal()}. That is correct for a background loop with no caller (see
+ * PrimaryRegistry#currentPrimaryTerminal()}. That is correct for a background loop with no caller (see
  * {@code dev.ltms.fleet.msg.LeadHeartbeatLoop}), but wrong here and a violation of this project's
  * own charter invariant 3 — "identity comes from the connection, never an argument." This daemon
  * can hold more than one labelled lead tab (see {@code LeadLauncher}'s fleetd #359 two-reading
@@ -115,19 +127,17 @@ public final class LeadRollover {
 
     private static final Logger log = LoggerFactory.getLogger(LeadRollover.class);
 
-    /** Poll interval while waiting for the lead's pane to settle after {@code /clear}. */
-    static final long SETTLE_POLL_MS = 250;
+    /** Poll interval shared by every bounded wait in this class. */
+    static final long POLL_INTERVAL_MS = 250;
 
     /**
-     * How many consecutive not-yet-picked-up polls {@link #waitForClearPickupAndSettle} allows
-     * before releasing rather than wedging the roll — the same constant and the same
-     * release-not-wedge choice {@link dev.ltms.fleet.inject.Injector} already makes for its own
-     * post-turn {@code /clear} housekeeping (fleetd #306). <strong>This bounds the number of
-     * consecutive polls, not the number of nudges:</strong> the first {@code PICKUP_GRACE_POLLS - 1}
-     * of those polls each send a nudge, and the {@code PICKUP_GRACE_POLLS}th releases instead of
-     * nudging again — so 8 polls produce 7 nudges, not 8.
+     * How long {@link #waitUntilPaneGone} polls {@link WorkspaceControl#locatePane} before giving
+     * up on ever seeing the old pane disappear. Not configurable: once {@link #endOldSession} has
+     * closed the pane (and, usually, its tab), herdr dropping the pane from its own bookkeeping is
+     * expected to show up within one or two polls, not on an operator-tunable timescale the way a
+     * CLI boot is.
      */
-    static final int PICKUP_GRACE_POLLS = 8;
+    static final int PANE_DEATH_TIMEOUT_SECONDS = 10;
 
     /**
      * One request opened by {@link #open}, pending its {@link #confirm} (or {@link #cancel}).
@@ -139,9 +149,17 @@ public final class LeadRollover {
      *                     calling lead's workspace, before storing it here; see this class's
      *                     javadoc. This is the value the MCP layer hands back to the lead as
      *                     "write your file here", so callers may rely on it always being absolute.
+     * @param rolloverKey  the key {@link #confirm}'s single-flight claim is taken under. {@link
+     *                     #open} resolves this once, here, from {@code leadTerminal} while the
+     *                     calling lead is certainly still live: the lead's configured name, or
+     *                     {@code leadTerminal} itself when no name resolves. Carried rather than
+     *                     recomputed at release time, because by the time a roll's continuation
+     *                     releases its claim the OLD terminal may no longer resolve to any name at
+     *                     all — recomputing there would release a different key than the one the
+     *                     claim was taken under.
      */
     public record PendingRollover(String token, String leadTerminal, String handoverPath,
-                                   long requestedAtMillis) {}
+                                   long requestedAtMillis, String rolloverKey) {}
 
     /** Which check refused a {@link #confirm} call, named so a caller can act on it. */
     public enum RefusalReason {
@@ -164,15 +182,21 @@ public final class LeadRollover {
          * The handover file's modified time is not after {@link #open}'s request timestamp, or is
          * older than {@code maxDocAgeSeconds}.
          */
-        HANDOVER_STALE
+        HANDOVER_STALE,
+        /**
+         * This lead already has a roll running: an earlier {@link #confirm} call claimed its
+         * single-flight key (see {@link PendingRollover#rolloverKey}) and that roll's continuation
+         * has not released it yet. {@code detail} names the key and the token that holds the claim.
+         */
+        ROLL_ALREADY_RUNNING
     }
 
     /**
      * The outcome of a {@link #confirm} call. {@link #approved()} means every gate passed and the
-     * roll has been handed to a one-shot continuation — <strong>not</strong> that the pane has been
-     * cleared; the continuation may still be waiting for the calling turn to end when this returns.
-     * Whether the deferred roll itself later goes on to clear the pane, refuse for never going
-     * idle, or refuse for never re-settling after {@code /clear} is logged only (see this class's
+     * roll has been handed to a one-shot continuation — <strong>not</strong> that the lead has
+     * already been replaced; the continuation may still be waiting for the calling turn to end when
+     * this returns. Whether the deferred roll itself later goes on to tear the old pane down and
+     * relaunch the lead, or refuses at any of its own steps, is logged only (see this class's
      * javadoc) — there is deliberately no synchronous caller left by that point to hand a result to.
      */
     public record RollDecision(boolean accepted, RefusalReason reason, String detail) {
@@ -208,7 +232,7 @@ public final class LeadRollover {
 
     /**
      * What is known about one token, right now — the answer {@link #status} gives. Distinguishes
-     * three terminal outcomes an approved roll can finish with, one in-flight outcome for a roll
+     * five terminal outcomes an approved roll can finish with, one in-flight outcome for a roll
      * that has been approved but has not finished yet, and two answers for a token that names no
      * active work at all: still pending confirmation, or nothing known about this token at all.
      */
@@ -231,41 +255,64 @@ public final class LeadRollover {
          * #status} could wrongly answer {@link #UNKNOWN} ("nothing was ever requested") for a roll
          * that is, in fact, actively running. This is not sticky: the deferred continuation
          * overwrites this same entry with a terminal state ({@link #ROLLED}, {@link
-         * #TURN_NEVER_SETTLED}, {@link #CLEAR_NEVER_SETTLED}, or {@link #FAILED}) once it finishes
-         * — including by throwing, which fleetd #615's catch in {@link #runRollover} now turns into
-         * {@link #FAILED} instead of leaving this entry stuck forever.
+         * #TURN_NEVER_SETTLED}, {@link #OLD_PANE_NEVER_DIED}, {@link #RELAUNCH_FAILED}, {@link
+         * #RELAUNCH_NEVER_READY}, {@link #RELAUNCH_NOT_RECOGNISED}, or {@link #FAILED}) once it
+         * finishes — including by throwing, which {@link #runRollover}'s catch turns into {@link
+         * #FAILED} instead of leaving this entry stuck forever.
          */
         IN_PROGRESS,
         /**
-         * {@link #confirm} was approved and the deferred continuation completed the entire roll:
-         * the calling lead's turn settled, {@code /clear} was sent and settled, and {@code
-         * bootstrapText} was sent.
+         * {@link #confirm} was approved and the deferred continuation completed the entire roll: the
+         * calling lead's turn settled, the old pane was torn down and confirmed gone, a fresh lead
+         * was launched and recognised, and {@code bootstrapText} was sent to it.
          */
         ROLLED,
         /**
          * {@link #confirm} was approved, but the calling lead's own turn never reached a boundary
-         * (IDLE or DONE) within {@code turnSettleSeconds} — no {@code /clear} was ever sent, at
-         * all. This is the branch the fleetd #480 correction exists to make safe, and the one this
-         * status exists to make VISIBLE: before this, a lead that hit this case had no way to find
-         * out, and would carry on believing it was about to be replaced. See this class's javadoc.
+         * (IDLE or DONE) within {@code turnSettleSeconds} — the old pane was never touched at all.
+         * This is the state that makes a lead's own stuck turn VISIBLE: without it, a lead that hit
+         * this case would have no way to find out, and would carry on believing it was about to be
+         * replaced. See this class's javadoc.
          */
         TURN_NEVER_SETTLED,
         /**
-         * {@link #confirm} was approved and {@code /clear} was sent, but the pane never re-settled
-         * within {@code clearSettleSeconds} — {@code bootstrapText} was never sent.
+         * {@link #confirm} was approved and the calling lead's turn settled, the old pane was closed
+         * (and its tab, if it was the sole occupant), but {@link
+         * dev.ltms.fleet.herdr.WorkspaceControl#locatePane} kept reporting it as still present for
+         * the whole pane-death timeout. No relaunch was ever attempted, and {@code bootstrapText}
+         * was never sent.
          */
-        CLEAR_NEVER_SETTLED,
+        OLD_PANE_NEVER_DIED,
         /**
-         * fleetd #615: the deferred continuation threw a {@link RuntimeException} — most likely a
-         * {@link dev.ltms.fleet.herdr.HerdrException} out of one of the two unwrapped {@code
-         * agents.send} calls in {@link #runRollover} — and the continuation thread died with it.
-         * Before this state existed, that throw left {@link #outcomes} holding {@link #IN_PROGRESS}
-         * forever, because the production {@code continuationRunner} is a bare virtual thread with
-         * no uncaught-exception handler and nothing downstream of the throw ever ran to write a
-         * terminal outcome. {@code detail} names the exception, so a reader has something to act on
-         * — the same diagnostic style as {@link #TURN_NEVER_SETTLED} and {@link
-         * #CLEAR_NEVER_SETTLED}. The roll is dead at this point and does not retry itself; a stuck
-         * lead must {@link #open} a fresh request.
+         * The old pane was confirmed gone, but {@code LeadLauncher#relaunch} returned {@code null}
+         * — every launch attempt failed. {@code bootstrapText} was never sent, and no fresh terminal
+         * exists for this roll to have recognised.
+         */
+        RELAUNCH_FAILED,
+        /**
+         * A fresh lead was launched, but its pane never reached a real turn boundary ({@code IDLE}
+         * or {@code DONE}, never merely {@code BLOCKED}) within {@code relaunchReadySeconds} — the
+         * CLI never finished booting, or it stayed paused on a startup prompt. {@code bootstrapText}
+         * was never sent: typing into a pane that is not actually ready to accept input loses the
+         * keystrokes.
+         */
+        RELAUNCH_NEVER_READY,
+        /**
+         * A fresh lead was launched and its pane reached a real turn boundary, so {@code
+         * bootstrapText} WAS sent to it, but the terminal was never recognised as a live lead —
+         * present in the live-lead terminal map — within {@code relaunchReadySeconds}. The session
+         * itself is alive and bootstrapped; only the daemon's own bookkeeping has not caught up, and
+         * an operator should check why the tab was not recognised.
+         */
+        RELAUNCH_NOT_RECOGNISED,
+        /**
+         * The deferred continuation threw a {@link RuntimeException} and the continuation thread
+         * died with it. Without this state, that throw would leave {@link #outcomes} holding {@link
+         * #IN_PROGRESS} forever, because the production {@code continuationRunner} is a bare virtual
+         * thread with no uncaught-exception handler and nothing downstream of the throw ever runs to
+         * write a terminal outcome. {@code detail} names the exception, so a reader has something to
+         * act on. The roll is dead at this point and does not retry itself; a stuck lead must
+         * {@link #open} a fresh request.
          */
         FAILED,
         /**
@@ -286,6 +333,10 @@ public final class LeadRollover {
     public record RollStatus(RollState state, String detail) {}
 
     private final AgentControl agents;
+    /** Workspace/tab/pane control — used to tear down the old pane and confirm it is gone. */
+    private final WorkspaceControl spaces;
+    /** Starts the fresh lead that replaces the one this roll tears down. */
+    private final LeadLauncher launcher;
     private final Supplier<FleetConfig.LeadRollover> configSupplier;
     /**
      * Terminal id → that lead's configured workspace directory (their {@code
@@ -295,8 +346,24 @@ public final class LeadRollover {
      * daemon-cwd bug this parameter exists to fix.
      */
     private final Function<String, String> leadWorkspace;
+    /**
+     * Terminal id → that lead's configured name under {@code fleet.leaders}, or {@code null} when
+     * the terminal names no currently-recognised lead. {@link #open} calls this on the calling
+     * lead's own terminal, while it is certainly still live, to resolve {@link
+     * PendingRollover#rolloverKey}. The deferred continuation also calls this, on the OLD terminal,
+     * before tearing it down, so it knows which lead to pass to {@link LeadLauncher#relaunch} — by
+     * that point the live roster may no longer contain the old terminal, so this lookup can return
+     * {@code null} here even though {@link #open}'s earlier call against the same terminal did not.
+     */
+    private final Function<String, String> leadNameForTerminal;
+    /**
+     * The daemon's current terminal id → lead name map, read fresh on every poll. The deferred
+     * continuation polls this for the FRESH terminal {@link LeadLauncher#relaunch} returns, to
+     * learn when that terminal has been recognised as a live lead — see this class's javadoc.
+     */
+    private final Supplier<Map<String, String>> liveLeadTerminals;
     private final LongSupplier nowMillis;
-    private final Runnable settleSleeper;
+    private final Runnable pollSleeper;
     /**
      * Launches the post-{@code confirm()} continuation. Production uses a single unstarted virtual
      * thread per confirmed request — see this class's javadoc for why that is a single-shot task,
@@ -305,6 +372,15 @@ public final class LeadRollover {
      */
     private final Consumer<Runnable> continuationRunner;
     private final Map<String, PendingRollover> pending = new ConcurrentHashMap<>();
+    /**
+     * {@link PendingRollover#rolloverKey} → the token of the roll currently holding that key
+     * exclusive, for {@link #confirm}'s single-flight claim. {@link #confirm} claims an entry here
+     * with an atomic put-if-absent once every other gate has passed, refusing with {@link
+     * RefusalReason#ROLL_ALREADY_RUNNING} when a claim is already held; {@link #runRollover}
+     * releases it in a {@code finally}, on both the success and the thrown-exception path. A key
+     * absent from this map has no roll currently in flight for it.
+     */
+    private final Map<String, String> rollingByLead = new ConcurrentHashMap<>();
     /**
      * Finished tokens → what actually happened, for {@link #status}. Bounded by {@link
      * #OUTCOME_HISTORY_CAP}, oldest evicted first ({@code removeEldestEntry} on an insertion-order
@@ -323,29 +399,40 @@ public final class LeadRollover {
                 }
             });
 
-    /** Production constructor — wall clock, real sleep between settle polls, a real virtual thread. */
-    public LeadRollover(AgentControl agents, Supplier<FleetConfig.LeadRollover> configSupplier,
-                         Function<String, String> leadWorkspace) {
-        this(agents, configSupplier, leadWorkspace, System::currentTimeMillis,
-                () -> sleepUninterruptibly(SETTLE_POLL_MS),
+    /** Production constructor — wall clock, real sleep between polls, a real virtual thread. */
+    public LeadRollover(AgentControl agents, WorkspaceControl spaces, LeadLauncher launcher,
+                         Supplier<FleetConfig.LeadRollover> configSupplier,
+                         Function<String, String> leadWorkspace,
+                         Function<String, String> leadNameForTerminal,
+                         Supplier<Map<String, String>> liveLeadTerminals) {
+        this(agents, spaces, launcher, configSupplier, leadWorkspace, leadNameForTerminal,
+                liveLeadTerminals, System::currentTimeMillis,
+                () -> sleepUninterruptibly(POLL_INTERVAL_MS),
                 r -> Thread.ofVirtual().name("lead-rollover-continuation-").start(r));
     }
 
     /**
-     * Full constructor — an injectable wall-clock supplier, settle-poll sleeper, and continuation
-     * runner, for tests. {@code nowMillis} MUST be a wall-clock source (e.g. {@code
+     * Full constructor — an injectable wall-clock supplier, poll sleeper, and continuation runner,
+     * for tests. {@code nowMillis} MUST be a wall-clock source (e.g. {@code
      * System.currentTimeMillis()}), never {@code System.nanoTime()}: the freshness check compares
      * against a file's modified time, which only a wall clock is comparable to, and {@code
-     * nanoTime} freezes while the host sleeps (fleetd #386).
+     * nanoTime} freezes while the host sleeps.
      */
-    LeadRollover(AgentControl agents, Supplier<FleetConfig.LeadRollover> configSupplier,
-                 Function<String, String> leadWorkspace, LongSupplier nowMillis,
-                 Runnable settleSleeper, Consumer<Runnable> continuationRunner) {
+    LeadRollover(AgentControl agents, WorkspaceControl spaces, LeadLauncher launcher,
+                 Supplier<FleetConfig.LeadRollover> configSupplier,
+                 Function<String, String> leadWorkspace,
+                 Function<String, String> leadNameForTerminal,
+                 Supplier<Map<String, String>> liveLeadTerminals,
+                 LongSupplier nowMillis, Runnable pollSleeper, Consumer<Runnable> continuationRunner) {
         this.agents = agents;
+        this.spaces = spaces;
+        this.launcher = launcher;
         this.configSupplier = configSupplier;
         this.leadWorkspace = leadWorkspace;
+        this.leadNameForTerminal = leadNameForTerminal;
+        this.liveLeadTerminals = liveLeadTerminals;
         this.nowMillis = nowMillis;
-        this.settleSleeper = settleSleeper;
+        this.pollSleeper = pollSleeper;
         this.continuationRunner = continuationRunner;
     }
 
@@ -362,8 +449,9 @@ public final class LeadRollover {
     /**
      * The lead says it is ready to be replaced. Generates a token and records the resolved
      * handover path, this moment's wall-clock timestamp (the baseline {@link #confirm} checks the
-     * handover file's modified time against), and {@code leadTerminal} — only that exact terminal
-     * may later {@link #confirm} this token.
+     * handover file's modified time against), {@code leadTerminal} — only that exact terminal may
+     * later {@link #confirm} this token — and {@link PendingRollover#rolloverKey}, resolved here
+     * from {@code leadTerminal} while the calling lead is certainly still live.
      *
      * @param leadTerminal the calling lead's terminal id, resolved by the MCP layer from the
      *                     connection (see this class's javadoc) — never a client-supplied value
@@ -383,7 +471,9 @@ public final class LeadRollover {
         String token = UUID.randomUUID().toString();
         long requestedAt = nowMillis.getAsLong();
         String resolvedPath = resolveHandoverPath(cfg.handoverPath(), leadTerminal);
-        PendingRollover p = new PendingRollover(token, leadTerminal, resolvedPath, requestedAt);
+        String leadName = leadNameForTerminal.apply(leadTerminal);
+        String rolloverKey = (leadName == null || leadName.isBlank()) ? leadTerminal : leadName;
+        PendingRollover p = new PendingRollover(token, leadTerminal, resolvedPath, requestedAt, rolloverKey);
         pending.put(token, p);
         if (resolvedPath.equals(cfg.handoverPath())) {
             log.info("lead-rollover: open token={} lead={} handoverPath={} reason={}",
@@ -487,6 +577,15 @@ public final class LeadRollover {
             return docCheck;
         }
 
+        // Single-flight claim: atomic put-if-absent, taken only after every other gate has
+        // passed, so a refused confirm() never takes it. A non-null previous value means a
+        // different, still-running roll already holds this lead's claim.
+        String holder = rollingByLead.putIfAbsent(p.rolloverKey(), token);
+        if (holder != null) {
+            return RollDecision.refused(RefusalReason.ROLL_ALREADY_RUNNING,
+                    "lead '" + p.rolloverKey() + "' already has a roll running under token " + holder);
+        }
+
         // Record IN_PROGRESS BEFORE removing from `pending` — see RollState#IN_PROGRESS and
         // OUTCOME_HISTORY_CAP's javadoc. This ordering means `token` is written into `outcomes`
         // while it is STILL present in `pending`; status() checks `outcomes` first (see that
@@ -494,34 +593,48 @@ public final class LeadRollover {
         // remove-then-put ordering would leave in which the token is in neither map.
         outcomes.put(token, new RollStatus(RollState.IN_PROGRESS,
                 "confirm() approved this roll and handed it to the deferred continuation; it has "
-                        + "not finished yet — still waiting for the calling turn to settle, for "
-                        + "/clear to be sent and settle, or for bootstrapText to be sent"));
+                        + "not finished yet — still waiting for the calling turn to settle, for the "
+                        + "old pane to be torn down and confirmed gone, for the fresh lead to be "
+                        + "recognised, or for bootstrapText to be sent"));
         pending.remove(token);
         log.info("lead-rollover: confirmed token={} lead={} — roll scheduled once the calling turn ends",
                 token, callerTerminal);
-        continuationRunner.accept(() -> runRollover(p, cfg));
+        try {
+            continuationRunner.accept(() -> runRollover(p, cfg));
+        } catch (RuntimeException e) {
+            // continuationRunner can reject the hand-off itself (e.g. a bounded executor's
+            // RejectedExecutionException) before runRollover ever starts, so runRollover's own
+            // finally — the only other place that releases rollingByLead — never runs either.
+            // Release the claim here and overwrite the IN_PROGRESS entry with a terminal outcome,
+            // or this lead could never be rolled again and status() would report IN_PROGRESS
+            // forever for a roll that in fact never started.
+            log.warn("lead-rollover: continuationRunner rejected token={} lead={}: {} — the roll "
+                            + "never started; releasing its claim and reporting it as FAILED",
+                    token, callerTerminal, e.toString(), e);
+            rollingByLead.remove(p.rolloverKey(), token);
+            outcomes.put(token, new RollStatus(RollState.FAILED,
+                    "continuationRunner rejected this roll before it ever started: " + e.toString()
+                            + " — the roll never ran; open() a fresh rollover request"));
+        }
         return RollDecision.approved();
     }
 
     /**
      * The single-shot continuation {@link #confirm} hands to {@code continuationRunner}. Runs
      * entirely after {@link #confirm} has returned to its caller — see this class's javadoc for the
-     * four-step order. There is no result to return to by this point, so every outcome is logged
-     * only.
+     * full order. There is no result to return to by this point, so every outcome is logged only.
      *
-     * <p><strong>fleetd #615 — the whole body is wrapped in one {@code try}.</strong> The two {@code
-     * agents.send} calls below are not wrapped individually: {@code send} → {@code agentCall} →
-     * {@code herdr.call} can throw an unchecked {@link dev.ltms.fleet.herdr.HerdrException} (see
-     * {@code AgentControl.java}), and the production {@code continuationRunner} is a bare virtual
-     * thread with no uncaught-exception handler (see this class's public constructor). Before this
-     * fix, either throw killed the continuation thread silently, leaving the {@link
-     * RollState#IN_PROGRESS} entry {@link #confirm} wrote at hand-off stuck forever — {@link
-     * #status} had no way to tell a dead roll from one still genuinely running. The {@code catch}
-     * below is scoped to the method body rather than to each {@code send} call individually, so it
-     * also covers anything else added to this continuation later, not just today's two call sites —
-     * the same reasoning that put the write-a-terminal-outcome step at each of this method's other
-     * exits (see the {@link RollState#TURN_NEVER_SETTLED} and {@link RollState#CLEAR_NEVER_SETTLED}
-     * branches below) rather than inside the helpers that detect them.</p>
+     * <p><strong>The whole body is wrapped in one {@code try}.</strong> Several calls below —
+     * {@code agents.get}, {@code agents.close}, {@code agents.send} — can throw an unchecked {@link
+     * dev.ltms.fleet.herdr.HerdrException} (see {@code AgentControl.java}), and the production
+     * {@code continuationRunner} is a bare virtual thread with no uncaught-exception handler (see
+     * this class's public constructor). An uncaught throw would kill the continuation thread
+     * silently, leaving the {@link RollState#IN_PROGRESS} entry {@link #confirm} wrote at hand-off
+     * stuck forever — {@link #status} would have no way to tell a dead roll from one still
+     * genuinely running. The {@code catch} below is scoped to the method body rather than to each
+     * call individually, so it also covers every call in this continuation, not a fixed list of
+     * call sites — the same reasoning that put the write-a-terminal-outcome step at each of this
+     * method's other exits rather than inside the helpers that detect them.</p>
      *
      * <p>Only {@link RuntimeException} is caught, matching the local convention {@link
      * #waitUntilAtTurnBoundary} already set around its own {@code agents.status} call — not the
@@ -539,6 +652,12 @@ public final class LeadRollover {
                     "the roll's continuation threw " + e.toString() + " — the roll is dead and will "
                             + "not retry itself; check the daemon log for the stack trace, then open() "
                             + "a fresh rollover request"));
+        } finally {
+            // Release the single-flight claim on both the normal return and the thrown-exception
+            // path above — a release only on success would leave this lead unrollable forever
+            // after one failure. The conditional two-argument remove only clears the entry this
+            // roll itself holds, never a different roll's claim on the same key.
+            rollingByLead.remove(p.rolloverKey(), p.token());
         }
     }
 
@@ -546,53 +665,237 @@ public final class LeadRollover {
     private void runRolloverUnguarded(PendingRollover p, FleetConfig.LeadRollover cfg) {
         String lead = p.leadTerminal();
         long rollStartMillis = nowMillis.getAsLong();
+
         TurnSettleResult turnResult = waitUntilAtTurnBoundary(lead, cfg.turnSettleSeconds());
         if (!turnResult.settled()) {
-            // fleetd #494 follow-up: this line had the SAME defect as the /clear-timeout line below
-            // — cfg.turnSettleSeconds() is the CONFIGURED budget, not how long this wait actually
-            // ran. Print the measured elapsed time alongside it, labelled, exactly like the /clear
-            // path already does.
             log.warn("lead-rollover: pane {} did not reach a turn boundary (IDLE or DONE) after "
-                            + "confirm() — refusing to send /clear at all; the calling lead's own "
-                            + "turn is still live and clearing it now would destroy live context "
+                            + "confirm() — the old pane is never touched; the calling lead's own "
+                            + "turn is still live and tearing it down now would destroy live context "
                             + "(token={}, configured={}s elapsed={}ms)",
                     lead, p.token(), cfg.turnSettleSeconds(), turnResult.elapsedMillis());
             outcomes.put(p.token(), new RollStatus(RollState.TURN_NEVER_SETTLED,
                     "the calling lead's own turn never reached a boundary (IDLE or DONE) within "
                             + "turnSettleSeconds=" + cfg.turnSettleSeconds() + "s (measured elapsed="
-                            + turnResult.elapsedMillis() + "ms) — no /clear was ever sent. If this "
-                            + "keeps happening, raise turnSettleSeconds in fleetd.yaml"));
+                            + turnResult.elapsedMillis() + "ms) — the old pane was never touched. If "
+                            + "this keeps happening, raise turnSettleSeconds in fleetd.yaml"));
             return;
         }
 
-        // This deliberately bypasses Injector, exactly like ClaudeCodeLauncher#clearContext:
-        // /clear is housekeeping, not a delegated turn, and routing it through Injector wedges the
-        // pane forever (see this class's javadoc).
-        agents.send(lead, "/clear");
-        ClearSettleResult clearResult = waitForClearPickupAndSettle(lead, cfg.clearSettleSeconds());
-        if (!clearResult.settled()) {
-            // fleetd #494: cfg.clearSettleSeconds() is the CONFIGURED budget, not how long the wait
-            // actually ran — an operator reading only that number wrongly believes it is a measured
-            // duration. Print the measured elapsed time and nudge count alongside it, each labelled,
-            // so the two can be compared at a glance.
-            log.warn("lead-rollover: pane {} did not reach a turn boundary (IDLE or DONE) after "
-                            + "/clear — NOT sending bootstrapText (token={}, configured={}s "
-                            + "elapsed={}ms nudges={})",
-                    lead, p.token(), cfg.clearSettleSeconds(), clearResult.elapsedMillis(),
-                    clearResult.nudges());
-            outcomes.put(p.token(), new RollStatus(RollState.CLEAR_NEVER_SETTLED,
-                    "/clear was sent, but the pane never re-settled within clearSettleSeconds="
-                            + cfg.clearSettleSeconds() + "s (measured elapsed=" + clearResult.elapsedMillis()
-                            + "ms, nudges=" + clearResult.nudges() + ") — bootstrapText was never sent"));
+        // Captured once, here, and never re-resolved from `lead` again below: once the pane is
+        // closed there is nothing left for a terminal lookup to find.
+        Agent oldAgent = captureAgentWithRetry(lead);
+        String oldPaneId = oldAgent.paneId();
+        String leadName = leadNameForTerminal.apply(lead);
+
+        endOldSession(oldPaneId);
+        DeathResult deathResult = waitUntilPaneGone(oldPaneId);
+        if (!deathResult.gone()) {
+            log.warn("lead-rollover: old pane {} for lead {} was never confirmed gone after being "
+                            + "closed — not attempting a relaunch (token={}, timeout={}s "
+                            + "elapsed={}ms)",
+                    oldPaneId, lead, p.token(), PANE_DEATH_TIMEOUT_SECONDS, deathResult.elapsedMillis());
+            outcomes.put(p.token(), new RollStatus(RollState.OLD_PANE_NEVER_DIED,
+                    "the old pane was closed, but locatePane kept reporting it as still present "
+                            + "after a pane-death timeout=" + PANE_DEATH_TIMEOUT_SECONDS
+                            + "s (measured elapsed=" + deathResult.elapsedMillis() + "ms) — no "
+                            + "relaunch was attempted"));
             return;
         }
-        agents.send(lead, cfg.bootstrapTextFor(p.handoverPath()));
+
+        Agent newAgent = launcher.relaunch(leadName);
+        if (newAgent == null) {
+            log.warn("lead-rollover: relaunch of lead '{}' (old terminal {}) failed every attempt "
+                            + "— bootstrapText was never sent (token={})", leadName, lead, p.token());
+            outcomes.put(p.token(), new RollStatus(RollState.RELAUNCH_FAILED,
+                    "lead '" + leadName + "' could not be relaunched — every attempt failed; "
+                            + "bootstrapText was never sent"));
+            return;
+        }
+
+        ReadinessResult readinessResult = waitUntilPaneReady(newAgent.terminalId(),
+                cfg.relaunchReadySeconds());
+        if (!readinessResult.ready()) {
+            log.warn("lead-rollover: fresh pane for lead '{}' (terminal {}) never reached a real "
+                            + "turn boundary — bootstrapText was never sent (token={}, configured={}s "
+                            + "elapsed={}ms)",
+                    leadName, newAgent.terminalId(), p.token(), cfg.relaunchReadySeconds(),
+                    readinessResult.elapsedMillis());
+            outcomes.put(p.token(), new RollStatus(RollState.RELAUNCH_NEVER_READY,
+                    "fresh terminal " + newAgent.terminalId() + " never reached a real turn "
+                            + "boundary (IDLE or DONE) within relaunchReadySeconds="
+                            + cfg.relaunchReadySeconds() + "s (measured elapsed="
+                            + readinessResult.elapsedMillis() + "ms) — bootstrapText was never "
+                            + "sent"));
+            return;
+        }
+
+        IdentityResult identityResult = waitUntilRecognisedAsLead(newAgent.terminalId(),
+                cfg.relaunchReadySeconds());
+        agents.send(newAgent.terminalId(), cfg.bootstrapTextFor(p.handoverPath()));
+        if (!identityResult.ready()) {
+            log.warn("lead-rollover: fresh terminal {} for lead '{}' is alive and bootstrapped, but "
+                            + "was never recognised as a live lead — an operator should check why "
+                            + "the tab was not recognised (token={}, configured={}s elapsed={}ms)",
+                    newAgent.terminalId(), leadName, p.token(), cfg.relaunchReadySeconds(),
+                    identityResult.elapsedMillis());
+            outcomes.put(p.token(), new RollStatus(RollState.RELAUNCH_NOT_RECOGNISED,
+                    "bootstrapText was sent to fresh terminal " + newAgent.terminalId() + ", but "
+                            + "it was never recognised as a live lead within relaunchReadySeconds="
+                            + cfg.relaunchReadySeconds() + "s (measured elapsed="
+                            + identityResult.elapsedMillis() + "ms) — check why the tab was not "
+                            + "recognised"));
+            return;
+        }
+
         long rollElapsedMillis = nowMillis.getAsLong() - rollStartMillis;
-        log.info("lead-rollover: rolled token={} lead={} elapsedMs={}", p.token(), lead, rollElapsedMillis);
+        log.info("lead-rollover: rolled token={} oldLead={} newTerminal={} elapsedMs={}",
+                p.token(), lead, newAgent.terminalId(), rollElapsedMillis);
         outcomes.put(p.token(), new RollStatus(RollState.ROLLED,
-                "rolled successfully in " + rollElapsedMillis + "ms"));
+                "rolled successfully in " + rollElapsedMillis + "ms; new terminal="
+                        + newAgent.terminalId()));
     }
 
+    /** Attempts {@link #captureAgentWithRetry} makes before letting the failure propagate. */
+    static final int CAPTURE_RETRIES = 3;
+
+    /**
+     * {@link AgentControl#get} for {@code lead}, retried up to {@link #CAPTURE_RETRIES} times. The
+     * terminal-to-pane lookup it goes through can report a genuinely live agent as not found (see
+     * {@code AgentControl#agentCall}'s own re-resolve-once behaviour), and one such false negative
+     * must not abort an otherwise-healthy roll. The result is captured once by the caller and never
+     * looked up again — see this class's javadoc.
+     *
+     * @throws RuntimeException the last failure, if every attempt fails — {@link #runRollover}'s
+     *                           catch turns that into {@link RollState#FAILED}
+     */
+    private Agent captureAgentWithRetry(String lead) {
+        RuntimeException last = null;
+        for (int attempt = 1; attempt <= CAPTURE_RETRIES; attempt++) {
+            try {
+                return agents.get(lead);
+            } catch (RuntimeException e) {
+                last = e;
+                log.debug("lead-rollover: agents.get({}) failed on attempt {}/{}: {}",
+                        lead, attempt, CAPTURE_RETRIES, e.toString());
+                if (attempt < CAPTURE_RETRIES) {
+                    pollSleeper.run();
+                }
+            }
+        }
+        throw last;
+    }
+
+    /**
+     * End the old lead's session: close its pane, then close its tab only when the pane was that
+     * tab's sole occupant — the same pane-then-tab teardown {@code HerdrPeerLauncher#stop} uses for
+     * a member. An already-gone pane counts as success; any other {@code agents.close} failure
+     * propagates, so a genuinely failed teardown is never reported as done. A failing
+     * {@code spaces.closeTab} never propagates — by the time it runs the pane is already closed, so
+     * it is cosmetic tidying, not a real teardown failure.
+     */
+    private void endOldSession(String paneId) {
+        WorkspaceControl.PaneLocation loc = spaces.locatePane(paneId);
+        try {
+            agents.close(paneId);
+        } catch (HerdrException e) {
+            if (!isAlreadyGone(e)) {
+                throw e;
+            }
+            log.debug("lead-rollover: pane.close({}) ignored — already gone: {}", paneId, e.getMessage());
+        }
+        if (loc != null && loc.tabPaneCount() == 1) {
+            try {
+                spaces.closeTab(loc.tabId());
+            } catch (RuntimeException e) {
+                log.warn("lead-rollover: tab.close({}) failed — the pane is already torn down, so "
+                        + "continuing; the tab may need manual cleanup: {}", loc.tabId(), e.getMessage());
+            }
+        } else if (loc != null) {
+            log.debug("lead-rollover: not closing tab {} — it holds {} panes (not a dedicated lead "
+                    + "tab)", loc.tabId(), loc.tabPaneCount());
+        }
+    }
+
+    /** True when a herdr error means the target is already gone (safe to treat as done). */
+    private static boolean isAlreadyGone(HerdrException e) {
+        return e.code() != null && e.code().endsWith("_not_found");
+    }
+
+    /**
+     * Poll {@link WorkspaceControl#locatePane} for {@code paneId} until it reports {@code null}
+     * (the pane is gone) or {@link #PANE_DEATH_TIMEOUT_SECONDS} elapses. Deliberately never calls
+     * {@link AgentControl#status} and never reads the live-lead terminal map — both answer a
+     * different question (whether an AGENT is live, not whether this PANE still exists) and
+     * {@code locatePane} alone catches a {@link HerdrException} from the underlying {@code
+     * pane.get} and turns it into {@code null} — see this class's javadoc.
+     */
+    private DeathResult waitUntilPaneGone(String paneId) {
+        long startMillis = nowMillis.getAsLong();
+        long deadline = startMillis + TimeUnit.SECONDS.toMillis(PANE_DEATH_TIMEOUT_SECONDS);
+        while (nowMillis.getAsLong() < deadline) {
+            if (spaces.locatePane(paneId) == null) {
+                return new DeathResult(true, nowMillis.getAsLong() - startMillis);
+            }
+            pollSleeper.run();
+        }
+        return new DeathResult(false, nowMillis.getAsLong() - startMillis);
+    }
+
+    /** The measured outcome of {@link #waitUntilPaneGone}. */
+    private record DeathResult(boolean gone, long elapsedMillis) {}
+
+    /**
+     * Poll until {@code newTerminal}'s own pane reaches a real turn boundary ({@link
+     * AgentStatus#IDLE} or {@link AgentStatus#DONE}, never merely {@link AgentStatus#BLOCKED}) —
+     * the same exclusion {@link #waitUntilAtTurnBoundary} applies to the calling lead's own turn,
+     * applied here to the fresh one, so {@code bootstrapText} is never typed into a pane that has
+     * not actually finished booting — or {@code readySeconds} elapses. A failed status read
+     * degrades to "not yet ready" and is retried on the next poll.
+     */
+    private ReadinessResult waitUntilPaneReady(String newTerminal, int readySeconds) {
+        long startMillis = nowMillis.getAsLong();
+        long deadline = startMillis + TimeUnit.SECONDS.toMillis(readySeconds);
+        while (nowMillis.getAsLong() < deadline) {
+            AgentStatus status;
+            try {
+                status = agents.status(newTerminal);
+            } catch (RuntimeException e) {
+                log.debug("lead-rollover: status check failed while waiting for {} to be ready: {}",
+                        newTerminal, e.toString());
+                status = null;
+            }
+            if (status == AgentStatus.IDLE || status == AgentStatus.DONE) {
+                return new ReadinessResult(true, nowMillis.getAsLong() - startMillis);
+            }
+            pollSleeper.run();
+        }
+        return new ReadinessResult(false, nowMillis.getAsLong() - startMillis);
+    }
+
+    /** The measured outcome of {@link #waitUntilPaneReady}. */
+    private record ReadinessResult(boolean ready, long elapsedMillis) {}
+
+    /**
+     * Poll until {@code newTerminal} is present in {@link #liveLeadTerminals} or {@code
+     * readySeconds} elapses. This is bookkeeping, not a safety gate: the pane's own readiness (see
+     * {@link #waitUntilPaneReady}) is what decides whether {@code bootstrapText} is safe to send —
+     * a timeout here only means the daemon's own lead-discovery scan has not caught up yet.
+     */
+    private IdentityResult waitUntilRecognisedAsLead(String newTerminal, int readySeconds) {
+        long startMillis = nowMillis.getAsLong();
+        long deadline = startMillis + TimeUnit.SECONDS.toMillis(readySeconds);
+        while (nowMillis.getAsLong() < deadline) {
+            if (liveLeadTerminals.get().containsKey(newTerminal)) {
+                return new IdentityResult(true, nowMillis.getAsLong() - startMillis);
+            }
+            pollSleeper.run();
+        }
+        return new IdentityResult(false, nowMillis.getAsLong() - startMillis);
+    }
+
+    /** The measured outcome of {@link #waitUntilRecognisedAsLead}. */
+    private record IdentityResult(boolean ready, long elapsedMillis) {}
+
     /** Drop a pending request without rolling. @return whether a pending request existed for {@code token} */
     public boolean cancel(String token) {
         return pending.remove(token) != null;
@@ -682,15 +985,12 @@ public final class LeadRollover {
 
     /**
      * Poll {@link AgentControl#status} until {@code target} reports a real turn boundary — {@link
-     * AgentStatus#IDLE} or {@link AgentStatus#DONE} — bounded by {@code settleSeconds}. Used once by
-     * {@link #runRollover}, to wait for the CALLING turn's own pane to settle before {@code /clear}
-     * is ever sent at all — the {@code turnSettleSeconds} gate that makes this correction safe. The
-     * SECOND wait, after {@code /clear}, is {@link #waitForClearPickupAndSettle} instead (fleetd
-     * #489) — a plain boundary check is not enough there, because {@code /clear} starts no turn of
-     * its own, so this method would (wrongly) report "settled" on its very first poll whether or not
-     * {@code /clear} was actually picked up. A failed status read degrades to "not yet settled" and
-     * is retried on the next poll, the same posture {@code LeadHeartbeatLoop} and {@code
-     * HerdrPeerLauncher}'s readiness gate already take toward an unreadable status.
+     * AgentStatus#IDLE} or {@link AgentStatus#DONE} — bounded by {@code settleSeconds}. Used by
+     * {@link #runRollover} to wait for the CALLING turn's own pane to settle before the old pane is
+     * touched at all — the {@code turnSettleSeconds} gate that makes tearing it down safe. A failed
+     * status read degrades to "not yet settled" and is retried on the next poll, the same posture
+     * {@code LeadHeartbeatLoop} and {@code HerdrPeerLauncher}'s readiness gate already take toward
+     * an unreadable status.
      *
      * <p><strong>Deliberately not {@link AgentStatus#injectable()}.</strong> {@code injectable()}
      * answers the {@code Injector}'s question — "may I deliver a message without stepping on a live
@@ -698,16 +998,16 @@ public final class LeadRollover {
      * an approval prompt is safe to queue a message behind. This class asks a stricter question —
      * "has the turn actually ended" — and {@code BLOCKED} answers no: it is a live turn that is
      * merely paused, not one that has finished. Reusing {@code injectable()} here would let this
-     * wait fire {@code /clear} while the lead's own {@code confirm()}-calling turn is still live and
-     * paused on a prompt — exactly the live-context-destroying failure the {@code turnSettleSeconds}
-     * gate exists to prevent. Do not "simplify" this back to {@code injectable()}. ({@link
-     * #waitForClearPickupAndSettle} keeps the same exclusion of {@code BLOCKED}, for the same
-     * reason, on the second wait.)
+     * wait tear the old pane down while the lead's own {@code confirm()}-calling turn is still live
+     * and paused on a prompt — exactly the live-context-destroying failure {@code turnSettleSeconds}
+     * exists to prevent. Do not "simplify" this back to {@code injectable()}. ({@link
+     * #waitUntilPaneReady} applies the same exclusion of {@code BLOCKED} to the fresh lead's own
+     * turn.)
      *
      * @return a {@link TurnSettleResult} whose {@code settled()} is {@code true} once a real
      *         boundary was observed, {@code false} if {@code settleSeconds} elapses first.
      *         {@code elapsedMillis()} is a MEASURED value from the injected {@link #nowMillis}
-     *         clock, never the configured {@code settleSeconds} budget (fleetd #494 follow-up).
+     *         clock, never the configured {@code settleSeconds} budget.
      */
     private TurnSettleResult waitUntilAtTurnBoundary(String target, int settleSeconds) {
         long startMillis = nowMillis.getAsLong();
@@ -724,142 +1024,11 @@ public final class LeadRollover {
             if (status == AgentStatus.IDLE || status == AgentStatus.DONE) {
                 return new TurnSettleResult(true, nowMillis.getAsLong() - startMillis);
             }
-            settleSleeper.run();
+            pollSleeper.run();
         }
         return new TurnSettleResult(false, nowMillis.getAsLong() - startMillis);
     }
 
-    /**
-     * The measured outcome of {@link #waitUntilAtTurnBoundary} — fleetd #494 follow-up. The sibling
-     * of {@link ClearSettleResult} for the FIRST wait, which never nudges, so it carries no nudge
-     * count.
-     */
+    /** The measured outcome of {@link #waitUntilAtTurnBoundary}. */
     private record TurnSettleResult(boolean settled, long elapsedMillis) {}
-
-    /**
-     * The SECOND wait in {@link #runRollover} — after {@code /clear} has been sent, waits for it to
-     * settle, bounded by {@code settleSeconds}. <strong>fleetd #489 — the paste-race fix.</strong>
-     * {@code /clear} does not start a real turn of its own, so a pane with no submit race simply
-     * stays {@link AgentStatus#IDLE} the whole time: {@link #waitUntilAtTurnBoundary} would (wrongly)
-     * call that "settled" on its very first poll, whether or not the {@code /clear} Enter actually
-     * landed. That was Fault 1, measured live on 2026-09-12 — the second gate was a no-op, so a
-     * {@code bootstrapText} send followed immediately, racing Fault 2: {@link AgentControl#submit}'s
-     * own javadoc already records that the submit accompanying a delivery "can race the paste —
-     * especially right as the worker's TUI becomes interactive — leaving the text unsubmitted"
-     * (CB-113). Because {@code runRollover} deliberately bypasses {@code Injector} for {@code
-     * /clear} (see this class's javadoc), it inherited none of {@code Injector}'s nudging — so the
-     * lost {@code /clear} Enter sat in the input box and {@code bootstrapText} was typed right after
-     * it, landing as one concatenated line.
-     *
-     * <p>This method copies the pickup-nudge pattern {@link dev.ltms.fleet.inject.Injector} already
-     * ships for exactly this, on its own post-turn {@code /clear} housekeeping (fleetd #306; see
-     * {@code Injector.java:288-340} and {@code Injector.java:437-442}):
-     * <ul>
-     *   <li>an {@link AgentStatus#WORKING} sample means {@code /clear} was picked up as a real
-     *       turn;</li>
-     *   <li>until that happens, each poll that still reports {@link AgentStatus#IDLE} or {@link
-     *       AgentStatus#DONE} re-sends the submit keystroke ({@link AgentControl#submit}) to nudge
-     *       the raced Enter — for the first {@code PICKUP_GRACE_POLLS - 1} of {@link
-     *       #PICKUP_GRACE_POLLS} consecutive such polls (i.e. {@code PICKUP_GRACE_POLLS - 1}
-     *       nudges: 7, not 8, given {@code PICKUP_GRACE_POLLS = 8}). A second Enter on an empty
-     *       Claude Code prompt is a no-op, so repeating it is safe;</li>
-     *   <li>the {@code PICKUP_GRACE_POLLS}th consecutive such poll, with {@code WORKING} still never
-     *       observed, releases rather than wedges the roll instead of nudging again — the same
-     *       choice {@code Injector} makes — and returns {@code settled() == true} anyway, logged at
-     *       {@code warn} with the measured elapsed time (fleetd #494) so an operator can see which
-     *       path ran and how long it actually took;</li>
-     *   <li>once {@code WORKING} has been observed, nudging stops and this instead waits for a real
-     *       {@code working → IDLE/DONE} completion boundary before returning {@code true}.</li>
-     * </ul>
-     *
-     * <p><strong>{@link AgentStatus#BLOCKED} is deliberately excluded from both the nudge and the
-     * boundary check</strong> — the same reasoning as {@link #waitUntilAtTurnBoundary}'s own
-     * javadoc: a paused live turn is not a settled one, and re-sending Enter into an open approval
-     * prompt could wrongly answer it. A {@code BLOCKED} sample (or an unreadable/{@link
-     * AgentStatus#UNKNOWN} one) simply keeps this polling, with no nudge and no release, until either
-     * a real boundary is reached or {@code settleSeconds} runs out.
-     *
-     * <p>{@link AgentControl#submit} can itself throw; a {@link RuntimeException} from it is
-     * swallowed and logged at {@code debug}, exactly like {@code Injector.java:437-442} — a failed
-     * nudge must not abort the roll.
-     *
-     * @return a {@link ClearSettleResult} whose {@code settled()} is {@code true} once {@code
-     *         /clear} has settled, or once the nudge budget was exhausted with no pickup ever
-     *         observed (released rather than wedged); {@code false} if {@code settleSeconds} elapses
-     *         first — the caller must NOT send {@code bootstrapText} in that case, exactly as before
-     *         this fix. {@code elapsedMillis()} and {@code nudges()} are MEASURED values (from the
-     *         injected {@link #nowMillis} clock and an actual nudge count), never the configured
-     *         {@code settleSeconds} budget (fleetd #494).
-     */
-    private ClearSettleResult waitForClearPickupAndSettle(String target, int settleSeconds) {
-        long startMillis = nowMillis.getAsLong();
-        long deadline = startMillis + TimeUnit.SECONDS.toMillis(settleSeconds);
-        boolean pickedUp = false; // a WORKING sample has been observed since /clear was sent
-        int idlePollsAwaitingPickup = 0;
-        int nudges = 0;
-        while (nowMillis.getAsLong() < deadline) {
-            AgentStatus status;
-            try {
-                status = agents.status(target);
-            } catch (RuntimeException e) {
-                log.debug("lead-rollover: status check failed while waiting for {} to settle after "
-                        + "/clear: {}", target, e.toString());
-                status = null;
-            }
-            if (status == AgentStatus.WORKING) {
-                pickedUp = true;
-            } else if (status == AgentStatus.IDLE || status == AgentStatus.DONE) {
-                if (pickedUp) {
-                    // a real WORKING -> IDLE/DONE completion boundary
-                    return new ClearSettleResult(true, nowMillis.getAsLong() - startMillis, nudges);
-                }
-                if (++idlePollsAwaitingPickup >= PICKUP_GRACE_POLLS) {
-                    long elapsedMillis = nowMillis.getAsLong() - startMillis;
-                    // fleetd #494: this release trades a possibly-unsubmitted /clear for progress
-                    // instead of wedging the roll — that trade is deliberate and stays. But it is
-                    // also exactly the case that reported false success in the real incident (the
-                    // whole roll "succeeded" after 438ms of a 20s budget), so raise it to WARN and
-                    // print the MEASURED elapsed time next to the target pane, not just the count.
-                    //
-                    // fleetd #494 follow-up (2nd pass): BOTH numbers in this line must come from
-                    // the loop's own counters, never from the PICKUP_GRACE_POLLS constant.
-                    // `idlePollsAwaitingPickup` and `nudges` each have exactly one write site in
-                    // this loop, on the same branch, so on this branch they cannot differ from
-                    // PICKUP_GRACE_POLLS / PICKUP_GRACE_POLLS - 1 today — no test can prove the
-                    // difference on this line, and printing the counters does not change that.
-                    // What it does buy: one source of truth instead of two, so a later change to
-                    // the loop (an early return, a second increment site, a different exit
-                    // condition) cannot leave this message reporting a number the loop no longer
-                    // produces. The place where `nudges` genuinely varies with the run — and is
-                    // covered by a test that can tell it apart from a constant — is the
-                    // /clear-timeout warn in runRollover, which prints clearResult.nudges().
-                    log.warn("lead-rollover: /clear on {} was never observed as WORKING after {} "
-                                    + "consecutive IDLE/DONE polls ({} of those were nudged) — "
-                                    + "releasing rather than wedging the roll (elapsed={}ms)",
-                            target, idlePollsAwaitingPickup, nudges, elapsedMillis);
-                    return new ClearSettleResult(true, elapsedMillis, nudges);
-                }
-                try {
-                    agents.submit(target); // nudge a raced Enter (CB-113) so /clear actually submits
-                } catch (RuntimeException e) {
-                    log.debug("lead-rollover: resubmit to {} failed (will retry next poll): {}",
-                            target, e.getMessage());
-                } finally {
-                    nudges++; // an attempted nudge, whether or not the submit call itself threw
-                }
-            }
-            // AgentStatus.BLOCKED or UNKNOWN (or an unreadable status, above): neither a pickup
-            // signal nor a boundary — keep polling without nudging or releasing.
-            settleSleeper.run();
-        }
-        return new ClearSettleResult(false, nowMillis.getAsLong() - startMillis, nudges);
-    }
-
-    /**
-     * The measured outcome of {@link #waitForClearPickupAndSettle} — fleetd #494. Carries the
-     * MEASURED elapsed time (from the injected {@link #nowMillis} clock) and nudge count alongside
-     * the settle/timeout decision, so callers can log them instead of the configured budget, which
-     * is not how long the wait actually ran.
-     */
-    private record ClearSettleResult(boolean settled, long elapsedMillis, int nudges) {}
 }

fleetd/src/main/java/dev/ltms/fleet/mcp/ConnectionIdentity.java

@@ -5,13 +5,13 @@ import dev.ltms.fleet.herdr.PaneLocator;
 /**
  * Resolves <em>who is calling</em> an MCP tool from the connection alone — the anti-spoofing
  * identity model of the MCP contract. It ties the connection's loopback peer PID (from the OS)
- * to a herdr agent pane (from herdr), yielding the caller's worker {@code terminal_id}. A caller
- * that maps to no worker pane — the primary, or an off-host client — resolves to {@code null}.
+ * to a herdr agent pane (from herdr), yielding that pane's {@code terminal_id}. A connection that
+ * maps to no pane resolves to {@code null}; this class assigns no role to either outcome — {@link
+ * dev.ltms.fleet.auth.CallerResolver} does that.
  *
  * <p>Both sources are authoritative and unforgeable: the OS reports the real connecting PID, and
- * herdr owns the PID→pane mapping. A worker cannot claim to be another worker, nor the primary.
- * Single-host only (the herd shares the {@code fleetd} host); the token path is the split-host
- * fallback.
+ * herdr owns the PID→pane mapping, so a caller cannot claim to be at another pane. Single-host
+ * only (the herd shares the {@code fleetd} host); the token path is the split-host fallback.
  */
 public final class ConnectionIdentity {
 
@@ -46,9 +46,10 @@ public final class ConnectionIdentity {
     }
 
     /**
-     * The caller resolved from the connection: its worker {@code terminal} (or {@code null} for the
-     * primary / an off-host client), its {@code pid} (or {@code -1} if not resolvable), and whether
-     * the pane scan behind {@code terminal} ran to completion ({@link #scanComplete}).
+     * The caller resolved from the connection: the {@code terminal} of the pane it connects from
+     * (or {@code null} when the connection maps to no pane), its {@code pid} (or {@code -1} if not
+     * resolvable), and whether the pane scan behind {@code terminal} ran to completion
+     * ({@link #scanComplete}).
      */
     public record Caller(String terminal, long pid, boolean scanComplete) {
 
@@ -87,8 +88,8 @@ public final class ConnectionIdentity {
     }
 
     /**
-     * The calling worker's {@code terminal_id}, or {@code null} if the caller is not a known
-     * on-host worker (treat as the primary).
+     * The terminal id of the pane the caller connects from, or {@code null} if the connection
+     * maps to no pane.
      */
     public String callerTerminal(String remoteAddr, int remotePort) {
         return resolve(remoteAddr, remotePort).terminal();

fleetd/src/main/java/dev/ltms/fleet/mcp/PrimaryRegistry.java

@@ -6,6 +6,7 @@ import org.slf4j.LoggerFactory;
 import java.util.Optional;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.atomic.AtomicReference;
+import java.util.function.Function;
 
 /**
  * Single-slot, thread-safe registry for the primary's herdr {@code terminal_id}.
@@ -18,13 +19,25 @@ import java.util.concurrent.atomic.AtomicReference;
  * <p>The push loop ({@code ReplyPushLoop}) uses {@link #isKnown()} to decide
  * whether active nudging is possible; an empty registry means the primary is
  * off-host or non-herdr and delivery falls back to pull.
+ *
+ * <p><strong>A learned terminal can go stale; a configured lead's name cannot.</strong> A lead that
+ * is rolled (a fresh pane replacing the old one) keeps its name but gets a new {@code terminal_id}.
+ * So every terminal this class learns — the singleton and each per-target delegation — is recorded
+ * together with the delegating lead's name, when the caller carries one. {@link
+ * #currentPrimaryTerminal()} and {@link #nudgeTargetFor(String)} resolve that name back to a
+ * terminal through the live {@code currentTerminalForName} lookup before falling back to the
+ * terminal that was actually recorded. A caller with no name (an unnamed primary, an architect, a
+ * collaborator — none of those are leads a lookup keyed on lead names can resolve) is tracked by
+ * terminal alone, exactly as before this indirection existed.
  */
 public final class PrimaryRegistry {
 
     private static final Logger log = LoggerFactory.getLogger(PrimaryRegistry.class);
 
     private final AtomicReference<String> terminal = new AtomicReference<>();
+    private final AtomicReference<String> primaryName = new AtomicReference<>();
     private final boolean pinned;
+    private final Function<String, String> currentTerminalForName;
 
     /**
      * CB-532: worker terminal → the lead that delegated to it. The single slot above answers "who is
@@ -33,12 +46,32 @@ public final class PrimaryRegistry {
      * other lead's delegations. This map answers the question that actually matters — "who is
      * waiting on THIS worker" — and is what lets {@code primary.terminal} be retired.
      */
-    private final ConcurrentHashMap<String, String> leadByTarget = new ConcurrentHashMap<>();
+    private final ConcurrentHashMap<String, Delegation> leadByTarget = new ConcurrentHashMap<>();
+
+    /** A recorded delegator: the terminal learned from call traffic, and its name, if it has one. */
+    private record Delegation(String terminal, String name) {
+    }
 
     /**
      * @param pinnedTerminal an optional pinned terminal from config ({@code null}/blank = unpinned)
      */
     public PrimaryRegistry(String pinnedTerminal) {
+        this(pinnedTerminal, name -> null);
+    }
+
+    /**
+     * As above, with a live {@code lead name → current terminal} lookup — normally the inverse of
+     * the same {@code terminal_id → lead name} supplier {@code CallerResolver} and the lead-tab
+     * scan already read. A lookup that cannot place a name (it is not a currently recognised lead,
+     * or no lookup is wired) returns {@code null}, and every resolution here falls back to the
+     * terminal that was actually recorded.
+     *
+     * @param pinnedTerminal         an optional pinned terminal from config ({@code null}/blank =
+     *                               unpinned)
+     * @param currentTerminalForName lead name → its current terminal, or {@code null} if that name
+     *                               is not a currently recognised lead
+     */
+    public PrimaryRegistry(String pinnedTerminal, Function<String, String> currentTerminalForName) {
         if (pinnedTerminal != null && !pinnedTerminal.isBlank()) {
             this.terminal.set(pinnedTerminal);
             this.pinned = true;
@@ -46,19 +79,31 @@ public final class PrimaryRegistry {
         } else {
             this.pinned = false;
         }
+        this.currentTerminalForName = currentTerminalForName != null ? currentTerminalForName : name -> null;
     }
 
     /**
-     * Record a terminal_id. No-op when:
+     * Record a terminal_id, with no lead name. No-op when:
      * <ul>
      *   <li>the registry is pinned (config override),
      *   <li>{@code terminalId} is {@code null} or blank (non-herdr caller).
      * </ul>
      */
     public void record(String terminalId) {
+        record(terminalId, null);
+    }
+
+    /**
+     * As {@link #record(String)}, additionally recording the caller's name — present for a
+     * configured lead, {@code null} for an unnamed primary. The name is what lets {@link
+     * #currentPrimaryTerminal()} keep nudging the same lead across a roll even though its terminal
+     * changed.
+     */
+    public void record(String terminalId, String name) {
         if (pinned) return;
         if (terminalId == null || terminalId.isBlank()) return;
         String prev = terminal.getAndSet(terminalId);
+        primaryName.set(blankToNull(name));
         if (prev == null) {
             log.debug("primary terminal learned: {}", terminalId);
         } else if (!prev.equals(terminalId)) {
@@ -67,7 +112,8 @@ public final class PrimaryRegistry {
     }
 
     /**
-     * Record that {@code leadTerminal} owns the accepted delegation of worker {@code target} (CB-532).
+     * Record that {@code leadTerminal} owns the accepted delegation of worker {@code target}
+     * (CB-532), with no lead name.
      *
      * <p>Called from the {@code MessageService} accepted-delivery hook — only after a send has won
      * the session's send lock and queued delivery — where both halves are known (CB-548). It is
@@ -77,10 +123,19 @@ public final class PrimaryRegistry {
      * lead that most recently delegated to it, which is the one waiting.
      */
     public void recordDelegation(String target, String leadTerminal) {
+        recordDelegation(target, leadTerminal, null);
+    }
+
+    /**
+     * As {@link #recordDelegation(String, String)}, additionally recording the delegating lead's
+     * name when the caller carries one. See {@link #record(String, String)} for why the name
+     * matters.
+     */
+    public void recordDelegation(String target, String leadTerminal, String leadName) {
         if (target == null || target.isBlank() || leadTerminal == null || leadTerminal.isBlank()) {
             return;
         }
-        leadByTarget.put(target, leadTerminal);
+        leadByTarget.put(target, new Delegation(leadTerminal, blankToNull(leadName)));
     }
 
     /** Forget a worker's delegating lead — call on release, so a torn-down session leaks nothing. */
@@ -99,19 +154,59 @@ public final class PrimaryRegistry {
      * recorded delegation there is no right answer, so this returns empty rather than guessing —
      * delivery degrades to pull, which is exactly what the durable inbox is for, instead of
      * interrupting the wrong lead with someone else's result.
+     *
+     * <p>A delegation recorded with a name is resolved to that lead's <em>current</em> terminal
+     * first — see {@link #currentTerminalForName} — so a lead that has since been rolled is still
+     * reachable here, not just the pane that delegated the work originally.
      */
     public Optional<String> nudgeTargetFor(String target) {
-        String lead = target == null ? null : leadByTarget.get(target);
-        return lead != null ? Optional.of(lead) : Optional.ofNullable(terminal.get());
+        Delegation delegation = target == null ? null : leadByTarget.get(target);
+        if (delegation != null) {
+            return Optional.of(resolveCurrent(delegation.terminal(), delegation.name()));
+        }
+        return currentPrimaryTerminal();
     }
 
-    /** The known primary terminal, or empty if not yet learned (and not pinned). */
+    /**
+     * The known primary terminal, or empty if not yet learned (and not pinned) — the raw value as
+     * it was recorded, with no attempt to resolve a named lead's current pane. Callers that need a
+     * nudge destination which survives a lead roll want {@link #currentPrimaryTerminal()} instead.
+     */
     public Optional<String> primaryTerminal() {
         return Optional.ofNullable(terminal.get());
     }
 
+    /**
+     * The terminal to nudge for the singleton primary right now: the recorded name resolved to its
+     * current terminal when one was recorded and is still a recognised lead, otherwise the terminal
+     * that was actually recorded — empty only when nothing has been learned or pinned at all.
+     */
+    public Optional<String> currentPrimaryTerminal() {
+        String learned = terminal.get();
+        if (learned == null) {
+            return Optional.empty();
+        }
+        return Optional.of(resolveCurrent(learned, primaryName.get()));
+    }
+
     /** {@code true} once a terminal has been recorded (or was pinned at construction). */
     public boolean isKnown() {
         return terminal.get() != null;
     }
+
+    /**
+     * {@code learnedTerminal}, unless {@code name} is non-null and {@code currentTerminalForName}
+     * currently places that name at a different, live terminal — in which case the live one wins.
+     */
+    private String resolveCurrent(String learnedTerminal, String name) {
+        if (name == null) {
+            return learnedTerminal;
+        }
+        String current = currentTerminalForName.apply(name);
+        return current != null && !current.isBlank() ? current : learnedTerminal;
+    }
+
+    private static String blankToNull(String s) {
+        return s == null || s.isBlank() ? null : s;
+    }
 }

fleetd/src/main/java/dev/ltms/fleet/member/CompositePeerLauncher.java

@@ -445,27 +445,6 @@ public final class CompositePeerLauncher implements PeerLauncher {
                         + " distinct candidate(s): " + String.join(", ", unreachable));
     }
 
-    /**
-     * Refuse an explicit-profile spawn when the profile is at its {@code maxLoad} cap.
-     *
-     * <p>maxLoad is a documented, unconditional capacity limit (see {@code FleetConfig.Profile#maxLoad}),
-     * and the charter makes explicit-profile spawns the normal path — so enforcing it only in placement
-     * ({@code PlacementPolicyUtil}, package-private, hence not linked) would leave the cap dead config
-     * on every call that names a profile. Same rule as placement: {@code live >= cap} is at capacity.
-     *
-     * <p>Deliberately no fallback to another profile: the caller named {@code profile} for a cost/model
-     * reason, and silently re-routing a paid-tier (subscription) request elsewhere is worse than
-     * refusing it. A caller that wants placement should omit the profile and let the policy pick.
-     *
-     * <p>Known TOCTOU limitation — documented, not fixed. {@link #liveCount} is read outside any lock and
-     * {@code SessionManager} registers a session only after {@code launcher.spawn} returns, so two
-     * genuinely concurrent spawns can both pass this check. The race already exists on the placement
-     * path. Closing it needs slot reservation in the registry; serializing spawn here would block on
-     * the readiness gate and is a far worse trade.
-     *
-     * @param profile the profile the caller explicitly named
-     * @throws PlacementException when the profile is at capacity
-     */
     /**
      * Refuse an explicit-profile spawn whose credential is quarantined (CB-578 stage B): a prior
      * {@code BACKEND_EXHAUSTED} classification on this profile, or on another profile sharing its
@@ -527,6 +506,27 @@ public final class CompositePeerLauncher implements PeerLauncher {
                 .collect(Collectors.toSet());
     }
 
+    /**
+     * Refuse an explicit-profile spawn when the profile is at its {@code maxLoad} cap.
+     *
+     * <p>maxLoad is a documented, unconditional capacity limit (see {@code FleetConfig.Profile#maxLoad}),
+     * and the charter makes explicit-profile spawns the normal path — so enforcing it only in placement
+     * ({@code PlacementPolicyUtil}, package-private, hence not linked) would leave the cap dead config
+     * on every call that names a profile. Same rule as placement: {@code live >= cap} is at capacity.
+     *
+     * <p>No fallback to another profile: the caller named {@code profile} for a cost/model
+     * reason, and silently re-routing a paid-tier (subscription) request elsewhere is worse than
+     * refusing it. A caller that wants placement should omit the profile and let the policy pick.
+     *
+     * <p>Known TOCTOU limitation — documented, not fixed. {@link #liveCount} is read outside any lock and
+     * {@code SessionManager} registers a session only after {@code launcher.spawn} returns, so two
+     * genuinely concurrent spawns can both pass this check. The race already exists on the placement
+     * path. Closing it needs slot reservation in the registry; serializing spawn here would block on
+     * the readiness gate and is a far worse trade.
+     *
+     * @param profile the profile the caller explicitly named
+     * @throws PlacementException when the profile is at capacity
+     */
     private void enforceMaxLoad(String profile) {
         // Absent config, or a config whose maxLoad normalized to null (ABSENT ⇒ unlimited at load),
         // means no cap — never cap what wasn't configured. Note "non-positive ⇒ unlimited" was true

fleetd/src/main/java/dev/ltms/fleet/member/EnvAllowListScrub.java

@@ -474,7 +474,6 @@ public final class EnvAllowListScrub {
         }
     }
 
-    /** Best-effort recursive delete; failures are swallowed — JVM-exit cleanup is the backstop. */
     /**
      * Remove generated directories left behind by an earlier daemon process.
      *
@@ -511,6 +510,7 @@ public final class EnvAllowListScrub {
         }
     }
 
+    /** Best-effort recursive delete; failures are swallowed — JVM-exit cleanup is the backstop. */
     static void deleteRecursively(Path dir) {
         if (dir == null || !Files.exists(dir)) {
             return;

fleetd/src/main/java/dev/ltms/fleet/member/HerdrPeerLauncher.java

@@ -6,6 +6,7 @@ import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
 import dev.ltms.fleet.herdr.HerdrClient;
 import dev.ltms.fleet.herdr.HerdrException;
+import dev.ltms.fleet.herdr.ResilientAgentLaunch;
 import dev.ltms.fleet.herdr.Tab;
 import dev.ltms.fleet.herdr.Workspace;
 import dev.ltms.fleet.herdr.WorkspaceControl;
@@ -67,16 +68,6 @@ public abstract class HerdrPeerLauncher implements PeerLauncher {
 
     private static final Logger log = LoggerFactory.getLogger(HerdrPeerLauncher.class);
 
-    /** herdr rejects a duplicate agent {@code name}; we retry a bumped name this many times. */
-    private static final int NAME_RETRIES = 8;
-
-    /**
-     * Retries for {@code agent.start} against a seed pane whose shell has not reached its prompt
-     * yet — {@code tab.create}/{@code pane.split} return as soon as the pane exists, and herdr
-     * refuses to start an agent in a pane that is not "an available shell" ({@code agent_pane_busy}).
-     */
-    private static final int SHELL_READY_RETRIES = 20;
-
     private final String namePrefix;                       // label prefix: naming + reap scheme
     private final AgentControl agents;
     private final WorkspaceControl spaces;
@@ -766,90 +757,26 @@ public abstract class HerdrPeerLauncher implements PeerLauncher {
         // Protocol 19 resolves the executable from the agent kind (== namePrefix here), so
         // argv[0] — the configured executable — is dropped and only the extra args are passed.
         List<String> args = argv.isEmpty() ? argv : argv.subList(1, argv.size());
-        checkPaneCommandFits(cfg, argv);
-        HerdrException last = null;
-        for (int attempt = 0; attempt < NAME_RETRIES; attempt++) {
-            long seq = nameSeq.incrementAndGet();
-            String name = namePrefix + "-" + cfg.profile() + "-" + nameNonce + "-" + seq;
-            try {
-                return new Started(startAwaitingShellPrompt(name, args, paneId), seq);
-            } catch (HerdrException e) {
-                if (!"agent_name_taken".equals(e.code())) throw e;
-                log.debug("peer name '{}' taken, retrying", name);
-                last = e;
-            }
+        try {
+            ResilientAgentLaunch.checkFits(cfg.profile(), argv);
+        } catch (ResilientAgentLaunch.TooLargeException e) {
+            throw new PeerUnreachableException(e.getMessage());
         }
-        throw last;
-    }
-
-    /**
-     * fleetd #220: herdr does not exec the launch command — it TYPES it into the pane as one line,
-     * and a pty line buffer holds only {@value #PANE_COMMAND_BYTE_LIMIT} bytes (BSD/macOS {@code
-     * MAX_CANON}). Everything past that byte is dropped. Nothing reports it: herdr answers "agent
-     * started", the backend exits on the mangled argument it was handed, the pane closes, and the
-     * only symptom is {@link #waitUntilInjectableOrThrow} timing out 20 seconds later with no
-     * reason. That is exactly how #214 broke every claude-code spawn — one 50-byte flag pushed a
-     * 978-byte command to 1028, and the tail that got cut was {@code --autocompact 250000}.
-     *
-     * <p>So measure it here and refuse, loudly and immediately, rather than spawn something that
-     * cannot work. The estimate is deliberately conservative: fleetd cannot see herdr's quoting, so
-     * every argument is charged its own bytes plus a separator and a quote pair. An over-estimate
-     * costs a clear error at a length that was already unsafe; an under-estimate would let the
-     * silent truncation back in.
-     *
-     * @throws PeerUnreachableException when the command cannot fit — the same failure the spawn
-     *                                  would have hit anyway, named at the point it is still
-     *                                  explainable
-     */
-    private void checkPaneCommandFits(FleetConfig.Profile cfg, List<String> argv) {
-        int bytes = 0;
-        String longest = null;
-        int longestBytes = 0;
-        for (String arg : argv) {
-            int argBytes = arg == null ? 0 : arg.getBytes(java.nio.charset.StandardCharsets.UTF_8).length;
-            bytes += argBytes + QUOTING_OVERHEAD_PER_ARG;
-            if (argBytes > longestBytes) {
-                longestBytes = argBytes;
-                longest = arg;
-            }
-        }
-        if (bytes <= PANE_COMMAND_BYTE_LIMIT) {
-            return;
-        }
-        String culprit = longest == null ? "<none>"
-                : longest.substring(0, Math.min(longest.length(), 60)) + (longest.length() > 60 ? "…" : "");
-        throw new PeerUnreachableException(
-                "launch command for profile " + cfg.profile() + " is about " + bytes + " bytes, over the "
-                        + PANE_COMMAND_BYTE_LIMIT + "-byte limit of the pane line herdr types it into. "
-                        + "The pty would drop the tail silently and the backend would exit on a mangled "
-                        + "argument. Longest argument is " + longestBytes + " bytes: " + culprit
-                        + " — move it off the command line (a file flag) or shorten it.");
+        long[] lastSeq = {0};
+        Agent agent = ResilientAgentLaunch.startUniquelyNamed(agents, namePrefix, args, paneId,
+                attempt -> {
+                    lastSeq[0] = nameSeq.incrementAndGet();
+                    return namePrefix + "-" + cfg.profile() + "-" + nameNonce + "-" + lastSeq[0];
+                },
+                ResilientAgentLaunch.NAME_RETRIES, ResilientAgentLaunch.SHELL_READY_RETRIES, sleeper);
+        return new Started(agent, lastSeq[0]);
     }
 
     /**
      * The pty line buffer herdr types a launch command into: BSD/macOS {@code MAX_CANON}. Not a
-     * fleetd choice and not configurable — see {@link #checkPaneCommandFits}.
+     * fleetd choice and not configurable — see {@link ResilientAgentLaunch#checkFits}.
      */
-    static final int PANE_COMMAND_BYTE_LIMIT = 1024;
-
-    /** Per-argument allowance for the separating space and a shell quote pair fleetd cannot see. */
-    private static final int QUOTING_OVERHEAD_PER_ARG = 3;
-
-    /** Start the agent into {@code paneId}, waiting out the seed shell's boot with the sleeper. */
-    private Agent startAwaitingShellPrompt(String name, List<String> args, String paneId) {
-        HerdrException busy = null;
-        for (int attempt = 0; attempt < SHELL_READY_RETRIES; attempt++) {
-            try {
-                return agents.start(name, namePrefix, args, paneId);
-            } catch (HerdrException e) {
-                if (!"agent_pane_busy".equals(e.code())) throw e;
-                log.debug("pane {} not at its shell prompt yet, retrying agent.start", paneId);
-                busy = e;
-                sleeper.run();
-            }
-        }
-        throw busy;
-    }
+    static final int PANE_COMMAND_BYTE_LIMIT = ResilientAgentLaunch.PANE_COMMAND_BYTE_LIMIT;
 
     // --- discovery + reap ----------------------------------------------------------------------
 

fleetd/src/main/java/dev/ltms/fleet/msg/LeadCoordLoop.java

@@ -2,6 +2,7 @@ package dev.ltms.fleet.msg;
 
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
+import dev.ltms.fleet.herdr.PromptBox;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
@@ -23,7 +24,8 @@ import java.util.function.Supplier;
  * <p><strong>Status-gated, exactly like {@link ReplyPushLoop}.</strong> A pane may only be injected
  * into at a turn boundary ({@link AgentStatus#injectable()} — idle, blocked or done); pasting into
  * a live turn corrupts it. So a tick that finds the lead busy simply does nothing and comes back
- * later.
+ * later. The same holds for a lead whose prompt box holds unsubmitted text ({@link PromptBox}) —
+ * delivering there would submit the operator's half-typed line along with the message.
  *
  * <p><strong>Ack only after delivery.</strong> A message is acked — removed from the broker — only
  * once {@link AgentControl#send} has actually put it in the pane. Anything not delivered (no lead
@@ -52,6 +54,7 @@ public final class LeadCoordLoop {
 
     private final LeadChannel channel;
     private final AgentControl agents;
+    private final PromptBox promptBox;
     private final Supplier<Map<String, String>> leads;
     private final ScheduledExecutorService scheduler;
     private final long intervalMs;
@@ -73,6 +76,7 @@ public final class LeadCoordLoop {
                          ScheduledExecutorService scheduler, long intervalMs) {
         this.channel = channel;
         this.agents = agents;
+        this.promptBox = new PromptBox(agents);
         this.leads = leads;
         this.scheduler = scheduler;
         this.intervalMs = intervalMs;
@@ -149,6 +153,11 @@ public final class LeadCoordLoop {
                     lead, status, held.size());
             return;
         }
+        if (!promptBox.clearToSubmit(lead)) {
+            log.debug("lead coordination: lead {} has unsubmitted text in its prompt box, holding {} message(s)",
+                    lead, held.size());
+            return;
+        }
         try {
             agents.send(lead, DELIVERY_FORMAT.formatted(msg.from(), msg.content()));
         } catch (RuntimeException e) {

fleetd/src/main/java/dev/ltms/fleet/msg/LeadHeartbeatLoop.java

@@ -2,6 +2,7 @@ package dev.ltms.fleet.msg;
 
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
+import dev.ltms.fleet.herdr.PromptBox;
 import dev.ltms.fleet.lead.LeadContextGauge;
 import dev.ltms.fleet.mcp.PrimaryRegistry;
 import dev.ltms.fleet.metrics.FleetMetrics;
@@ -33,7 +34,7 @@ import java.util.function.Supplier;
  * no such block it is never constructed, so upgrading the daemon cannot silently acquire a behaviour
  * that spends the operator's model subscription on its own initiative (constraint 1).
  *
- * <p>Four invariants keep it from becoming a runaway subscription burner:
+ * <p>Five invariants keep it from becoming a runaway subscription burner:
  * <ol>
  *   <li><b>Status-gated</b> — a {@code WORKING} lead is making progress and is never touched; only an
  *       injectable (idle/done/blocked) lead is even considered (constraint 2).</li>
@@ -45,6 +46,8 @@ import java.util.function.Supplier;
  *   <li><b>Never races {@link ReplyPushLoop}</b> — while that loop is actively nudging any target this
  *       loop stands down, so two competing injections never start two turns in the same pane
  *       (constraint 6).</li>
+ *   <li><b>Never submits the operator's draft</b> — a nudge is held while the lead's prompt box holds
+ *       unsubmitted text ({@link PromptBox}), because the delivery pastes and submits in one call.</li>
  * </ol>
  *
  * <p><b>fleetd #609 — context-high notice.</b> Optionally ({@code contextHighNudge}, opt-in like the
@@ -65,6 +68,7 @@ public final class LeadHeartbeatLoop {
 
     private final PrimaryRegistry primaryRegistry;
     private final AgentControl agents;
+    private final PromptBox promptBox;
     private final ReplyInbox inbox;
     private final Supplier<List<MemberSession>> roster;
     private final ReplyPushLoop pushLoop;
@@ -135,6 +139,7 @@ public final class LeadHeartbeatLoop {
                              boolean requireOperatorConfirm) {
         this.primaryRegistry = primaryRegistry;
         this.agents = agents;
+        this.promptBox = new PromptBox(agents);
         this.inbox = inbox;
         this.roster = roster;
         this.pushLoop = pushLoop;
@@ -187,7 +192,9 @@ public final class LeadHeartbeatLoop {
         /** Idle past the quiet period with nothing pending and the cap exhausted — stop until new state appears. */
         QUIET_DONE,
         /** {@link ReplyPushLoop} is actively nudging — stand aside rather than start a competing turn. */
-        STAND_DOWN
+        STAND_DOWN,
+        /** The lead's prompt box holds unsubmitted text — hold the nudge rather than submit that text. */
+        DRAFT_HELD
     }
 
     /**
@@ -307,12 +314,12 @@ public final class LeadHeartbeatLoop {
      * (mirroring {@link ReplyPushLoop#tick(String)}) so tests can drive it directly with a fake clock and a
      * fake {@link AgentControl} instead of racing the scheduler thread. */
     void tick() {
-        boolean leadKnown = primaryRegistry.primaryTerminal().isPresent();
+        boolean leadKnown = primaryRegistry.currentPrimaryTerminal().isPresent();
         FleetState fleet = snapshot(inbox, roster);
         AgentStatus status = AgentStatus.UNKNOWN;
         LeadContextGauge.Reading reading = LeadContextGauge.Reading.unknown();
         if (leadKnown) {
-            String leadTerminal = primaryRegistry.primaryTerminal().orElseThrow();
+            String leadTerminal = primaryRegistry.currentPrimaryTerminal().orElseThrow();
             try {
                 status = agents.status(leadTerminal);
             } catch (RuntimeException e) {
@@ -330,6 +337,7 @@ public final class LeadHeartbeatLoop {
                 idleSinceNanos == NOT_IDLE ? null : idleSinceNanos,
                 quietCount, status, pushLoop.isActive(), leadKnown, fleet,
                 reading.state(), contextNotified);
+        d = holdIfOperatorIsTyping(d);
         applyDecision(d);
         switch (d.action()) {
             case INJECT -> injectNudge(d, fleet, reading);
@@ -337,11 +345,33 @@ public final class LeadHeartbeatLoop {
                 countNudge("exhausted");
                 contextNotified = d.contextNotified();
             }
-            case WAIT_IDLE, LEAD_BUSY, STAND_DOWN -> contextNotified = d.contextNotified();
+            case WAIT_IDLE, LEAD_BUSY, STAND_DOWN, DRAFT_HELD -> contextNotified = d.contextNotified();
         }
         scheduleNext();
     }
 
+    /**
+     * Turn a decision to inject into {@link Action#DRAFT_HELD} when the lead's prompt box holds text
+     * the operator has not submitted. The pane read happens only for a decision that would otherwise
+     * send, so a busy or debouncing lead costs no extra herdr call.
+     *
+     * <p>The held decision carries this tick's idle window but the <em>pre-tick</em> quiet count and
+     * context latch: nothing reached the pane, so neither the quiet budget nor the one context notice
+     * per stretch may be spent on it.
+     */
+    private Decision holdIfOperatorIsTyping(Decision d) {
+        if (d.action() != Action.INJECT) {
+            return d;
+        }
+        var lead = primaryRegistry.currentPrimaryTerminal();
+        if (lead.isEmpty() || promptBox.clearToSubmit(lead.get())) {
+            return d;
+        }
+        log.debug("idle-heartbeat: lead {} has unsubmitted text in its prompt box, holding the nudge",
+                lead.get());
+        return new Decision(Action.DRAFT_HELD, d.idleSinceNanos(), quietCount, contextNotified);
+    }
+
     /**
      * Persist the idle/quiet state a decision returned, so the next tick starts from it.
      *
@@ -367,7 +397,7 @@ public final class LeadHeartbeatLoop {
         // itself should be built from. Otherwise a HIGH stretch that is still latched would never see the
         // notice at all, defeating the very check this fixes.
         String notice = contextNotice(contextHighNudge, reading, contextNotified, requireOperatorConfirm);
-        var lead = primaryRegistry.primaryTerminal();
+        var lead = primaryRegistry.currentPrimaryTerminal();
         boolean sent = lead.isPresent() && trySend(lead.get(), fleet.nudgeText() + notice, notice);
         // The latch becomes true only when all three hold: decide() chose to notify, a notice was
         // actually included in the text, and the send reached the pane without throwing. Whenever no

fleetd/src/main/java/dev/ltms/fleet/msg/MessageService.java

@@ -1,5 +1,6 @@
 package dev.ltms.fleet.msg;
 
+import dev.ltms.fleet.auth.Principal;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
 import dev.ltms.fleet.herdr.HerdrRouter;
@@ -11,6 +12,7 @@ import org.slf4j.LoggerFactory;
 
 import java.util.ArrayList;
 import java.util.List;
+import java.util.Objects;
 import java.util.UUID;
 import java.util.concurrent.CompletableFuture;
 import java.util.concurrent.CompletionException;
@@ -87,7 +89,7 @@ public final class MessageService {
         /**
          * The worker paused mid-turn to ask the primary a question (CB-205); {@code text} is the
          * question and {@code turnId} correlates the answer. Not terminal — the primary answers with
-         * {@link #answer(String, String, long)} and the turn resumes.
+         * {@link #answer(String, String, long, String)} and the turn resumes.
          */
         QUESTION,
         /** Timed out after the message was delivered — the worker is still working. */
@@ -120,10 +122,16 @@ public final class MessageService {
         /** Another send to this session was in flight for the whole window. */
         BUSY,
         /**
-         * An answer ({@link #answer(String, String, long)}) referenced a {@code turnId} that is no
-         * longer open — the worker's {@code fleet_ask} already timed out or was answered.
+         * An answer ({@link #answer(String, String, long, String)}) referenced a {@code turnId}
+         * that is no longer open — the worker's {@code fleet_ask} already timed out or was answered.
          */
-        STALE_TURN
+        STALE_TURN,
+        /**
+         * An answer ({@link #answer(String, String, long, String)}) named a {@code turnId} that is
+         * still open, but the answering caller is not the caller whose accepted delegation opened
+         * it. Distinct from {@link #STALE_TURN} so a refusal is never reported as a lapsed turn.
+         */
+        NOT_TURN_OWNER
     }
 
     /**
@@ -133,7 +141,7 @@ public final class MessageService {
      *                {@link Outcome#COMPLETED_UNREPLIED}), or the question for {@link Outcome#QUESTION},
      *                else {@code null}
      * @param turnId  correlation id for a {@link Outcome#QUESTION} (answered via
-     *                {@link #answer(String, String, long)}), else {@code null}
+     *                {@link #answer(String, String, long, String)}), else {@code null}
      */
     public record Reply(Outcome outcome, String text, String turnId) {
         /** A reply with no correlation id (the common terminal outcomes). */
@@ -275,10 +283,16 @@ public final class MessageService {
          * {@link #abandon}) can never match again regardless of this flag's value.
          */
         private volatile boolean askTimedOut;
+        /**
+         * The owner key of the caller whose {@code fleet_send{wait:false}} created this ticket, or
+         * {@code null} for the unnamed primary and overloads that do not record a caller.
+         */
+        private final String creatorOwner;
 
-        private Task(String ticket, String target, LongSupplier nowNanos) {
+        private Task(String ticket, String target, LongSupplier nowNanos, String creatorOwner) {
             this.ticket = ticket;
             this.target = target;
+            this.creatorOwner = creatorOwner;
             this.createdNanos = nowNanos.getAsLong();
             future.whenComplete((reply, ex) -> completedNanos = nowNanos.getAsLong());
         }
@@ -340,6 +354,14 @@ public final class MessageService {
      */
     private final ConcurrentHashMap<String, Boolean> queuedDeliveries = new ConcurrentHashMap<>();
     private final AtomicLong ticketSeq = new AtomicLong();
+    /**
+     * Minted once per {@code MessageService} instance and folded into every ticket id (see
+     * {@link #sendAsync(String, String, Runnable, Principal)}). {@link #ticketSeq} alone restarts at
+     * zero for every instance, so without this a ticket id can be reused across instances and
+     * resolve to an unrelated {@link Task} with no error; this nonce makes that impossible, because
+     * an id minted by one instance can never match the id space of another.
+     */
+    private final String ticketBootNonce = UUID.randomUUID().toString().substring(0, 6);
     private final ExecutorService asyncExecutor = Executors.newThreadPerTaskExecutor(
             Thread.ofVirtual().name("bridge-async-", 0).factory());
 
@@ -656,7 +678,7 @@ public final class MessageService {
             case TIMED_OUT_WORKING, TIMED_OUT_QUEUED, TIMED_OUT_UNCONFIRMED, BUSY -> "timeout";
             case WORKER_FAILED -> "failed";
             case BACKEND_EXHAUSTED -> "backend_exhausted";
-            case STALE_TURN, QUESTION -> null; // not a completed delegation
+            case STALE_TURN, QUESTION, NOT_TURN_OWNER -> null; // not a completed delegation
         };
     }
 
@@ -915,14 +937,17 @@ public final class MessageService {
 
     /**
      * Deliver {@code content} to {@code target} (a herdr {@code terminal_id}) and block until the
-     * worker replies via {@link Rendezvous} or {@code timeoutMillis} elapses.
+     * worker replies via {@link Rendezvous} or {@code timeoutMillis} elapses. {@code callerOwner}
+     * identifies the caller making this call and is recorded as the turn's owner. It is the only
+     * caller {@link #answer(String, String, long, String)} will
+     * later accept an answer from if the worker pauses mid-turn to ask.
      */
-    public Reply send(String target, String content, long timeoutMillis) {
-        return send(target, content, timeoutMillis, null);
+    public Reply send(String target, String content, long timeoutMillis, String callerOwner) {
+        return send(target, content, timeoutMillis, null, callerOwner);
     }
 
     /**
-     * As {@link #send(String, String, long)}, but with an accepted-delivery hook.
+     * As {@link #send(String, String, long, String)}, but with an accepted-delivery hook.
      *
      * <p>{@code onAccepted} is invoked exactly once, once this send has won {@code target}'s send
      * lock and so become the <em>accepted target turn</em> — it runs <em>before</em> delivery is
@@ -933,12 +958,13 @@ public final class MessageService {
      * acceptance means a concurrent sender that times out {@code BUSY} can never steal ownership it
      * never earned. {@code null} disables the hook.
      */
-    public Reply send(String target, String content, long timeoutMillis, Runnable onAccepted) {
-        return send(target, content, timeoutMillis, onAccepted, null);
+    public Reply send(String target, String content, long timeoutMillis, Runnable onAccepted, String callerOwner) {
+        return send(target, content, timeoutMillis, onAccepted, null, callerOwner);
     }
 
     /** Run a send, optionally stopping an async task that teardown already failed before acceptance. */
-    private Reply send(String target, String content, long timeoutMillis, Runnable onAccepted, Task task) {
+    private Reply send(String target, String content, long timeoutMillis, Runnable onAccepted, Task task,
+            String callerOwner) {
         long deadlineNanos = System.nanoTime() + timeoutMillis * 1_000_000L;
         ReentrantLock lock = sessionLocks.computeIfAbsent(target, _ -> new ReentrantLock());
 
@@ -958,7 +984,7 @@ public final class MessageService {
             // open race). Opening first also means a throwing onAccepted (fired before enqueue) or an
             // enqueue failure is safely closed by the finally below: nothing is left queued, and the
             // failed send leaves no stale waiter behind.
-            CompletableFuture<Rendezvous.Resolution> reply = rendezvous.open(target);
+            CompletableFuture<Rendezvous.Resolution> reply = rendezvous.open(target, Rendezvous.Owner.of(callerOwner));
             // CB-640: this send now owns target's delivery, so any earlier stranded-reply or
             // still-queued fact no longer describes the live state — clear both rather than let
             // them outlive the send that supersedes them.
@@ -1137,19 +1163,30 @@ public final class MessageService {
      * mid-turn (already picked up), so the answer flows back through its own open {@code fleet_ask}
      * call, not a new status-gated delivery. The forward waiter is opened <em>before</em> the worker
      * is unblocked so a reply that lands the instant it resumes is not lost.
+     *
+     * <p>{@code callerOwner} identifies the caller making this call. It is checked against the
+     * turn's recorded owner (the caller whose
+     * accepted delegation opened it, see {@link #send(String, String, long, String)} and
+     * {@link #sendAsync(String, String, Runnable, Principal)}) before anything else runs: a mismatch,
+     * including a turn with no owner on record at all, returns {@link Outcome#NOT_TURN_OWNER}
+     * without touching the rendezvous, the session lock, or any async task bookkeeping.
      */
-    public Reply answer(String turnId, String content, long timeoutMillis) {
+    public Reply answer(String turnId, String content, long timeoutMillis, String callerOwner) {
         String workerSession = rendezvous.askSession(turnId);
         if (workerSession == null) {
             return new Reply(Outcome.STALE_TURN, null); // the ask lapsed (timed out or already answered)
         }
+        Rendezvous.Owner owner = rendezvous.askOwner(turnId);
+        if (!Rendezvous.Owner.permits(owner, callerOwner)) {
+            return new Reply(Outcome.NOT_TURN_OWNER, null);
+        }
         long deadlineNanos = System.nanoTime() + timeoutMillis * 1_000_000L;
         ReentrantLock lock = sessionLocks.computeIfAbsent(workerSession, _ -> new ReentrantLock());
         if (!tryLock(lock, remainingMillis(deadlineNanos))) {
             return new Reply(Outcome.BUSY, null);
         }
         try {
-            CompletableFuture<Rendezvous.Resolution> reply = rendezvous.open(workerSession);
+            CompletableFuture<Rendezvous.Resolution> reply = rendezvous.open(workerSession, owner);
             // fleetd #575: this try used to open below, AFTER the Task lookup/registration and the
             // STALE_TURN early return that follows it — so that return was covered only by a
             // hand-rolled copy of the finally's own cleanup pair, not the finally itself. Widening the
@@ -1279,8 +1316,20 @@ public final class MessageService {
      * @return the ticket to poll for the eventual result
      */
     public String sendAsync(String target, String content, Runnable onAccepted) {
-        String ticket = "task-" + ticketSeq.incrementAndGet();
-        Task task = new Task(ticket, target, nowNanos);
+        return sendAsync(target, content, onAccepted, null);
+    }
+
+    /**
+     * As {@link #sendAsync(String, String, Runnable)}, recording {@code creator}'s owner key as this
+     * ticket's owner. The key is derived here from the resolved principal so callers cannot pass a
+     * terminal address where an owner identity is required.
+     *
+     * @return the ticket to poll for the eventual result
+     */
+    public String sendAsync(String target, String content, Runnable onAccepted, Principal creator) {
+        String ticket = "task-" + ticketBootNonce + "-" + ticketSeq.incrementAndGet();
+        String creatorOwner = creator == null ? null : creator.ownerKey();
+        Task task = new Task(ticket, target, nowNanos, creatorOwner);
         tasks.put(ticket, task);
         if (pushLoop != null) {
             // CB-588: task.future only ever completes on a terminal phase (DONE or a failure) — a
@@ -1311,7 +1360,7 @@ public final class MessageService {
         }
         asyncExecutor.submit(() -> {
             try {
-                Reply result = send(target, content, ASYNC_TIMEOUT_MS, onAccepted, task);
+                Reply result = send(target, content, ASYNC_TIMEOUT_MS, onAccepted, task, creatorOwner);
                 if (result.outcome() == Outcome.QUESTION) {
                     // Keep the accepted owner until answer() finishes it. markAsyncQuestion may run
                     // just after resolveQuestion wakes this thread.
@@ -1343,15 +1392,44 @@ public final class MessageService {
     }
 
     /**
-     * Snapshot the state of an async delegation. Returns {@code null} for an unknown/expired ticket;
-     * otherwise a {@link Phase#PENDING} view (with the live worker status as detail), a
-     * {@link Phase#DONE} view carrying the reply, or a {@link Phase#FAILED} view with the reason.
+     * As {@link #poll(String, String)}, but bypasses the ownership check entirely via
+     * {@link #INTERNAL_NO_OWNER_CHECK}. No production code calls this overload — it exists for
+     * tests that only need the ticket's state and have no caller identity to pass.
      */
     public TaskView poll(String ticket) {
+        return poll(ticket, INTERNAL_NO_OWNER_CHECK);
+    }
+
+    /**
+     * Whether {@code ticket} was created by the session identified by {@code callerOwner} — the
+     * same comparison {@link #poll(String, String)} itself makes, exposed so a caller can be
+     * granted read access to a ticket before it actually polls one. Returns {@code false} for an
+     * unknown or expired ticket.
+     */
+    public boolean isTicketOwnedBy(String ticket, String callerOwner) {
+        Task task = tasks.get(ticket);
+        return task != null && Objects.equals(callerOwner, task.creatorOwner);
+    }
+
+    /**
+     * Snapshot the state of an async delegation. Returns {@code null} for an unknown/expired ticket.
+     * Refuses a {@code callerOwner} that differs from the owner that created the ticket (see
+     * {@link #sendAsync(String, String, Runnable, Principal)}) with a {@link Phase#FAILED} view that
+     * carries no reply text. The unnamed primary's owner key is {@code null}, matched the same way
+     * as any other key — it reads a ticket another unnamed primary created, and is refused on a
+     * ticket a named caller created. Otherwise returns a {@link Phase#PENDING} view (with the live
+     * worker status as detail), a {@link Phase#DONE} view carrying the reply, or a
+     * {@link Phase#FAILED} view with the reason.
+     */
+    public TaskView poll(String ticket, String callerOwner) {
         Task task = tasks.get(ticket);
         if (task == null) {
             return null;
         }
+        if (!ownsTicket(task, callerOwner)) {
+            return new TaskView(ticket, Phase.FAILED, null, null,
+                    "forbidden: this ticket was created by a different session", null);
+        }
         CompletableFuture<Reply> f = task.future;
         if (!f.isDone()) {
             Reply question = task.question;
@@ -1387,6 +1465,30 @@ public final class MessageService {
         return new TaskView(ticket, Phase.FAILED, null, null, detail, null);
     }
 
+    /**
+     * Marker passed as {@code callerOwner} to bypass the ownership check entirely. No
+     * {@link Principal#ownerKey()} ever produces this value — every real key is either
+     * {@code null} (the unnamed primary) or prefixed with its role, such as {@code "worker:"} or
+     * {@code "leader:"}. {@link #poll(String)} passes it; {@link #pendingAsk} has no matching
+     * no-check overload, so this stays package-private for the test that drives the bypass
+     * directly.
+     */
+    static final String INTERNAL_NO_OWNER_CHECK = "internal:no-owner-check";
+
+    /**
+     * Whether {@code callerOwner} may read {@code task}'s state. {@code callerOwner} is matched
+     * against the task's recorded owner key by equality, including a {@code null} match — the
+     * unnamed primary's owner key is {@code null}, so it owns a ticket another unnamed primary
+     * created and nothing else, the same rule every other role follows. The only caller that
+     * reads any ticket is {@link #INTERNAL_NO_OWNER_CHECK}. This differs from
+     * {@link Rendezvous.Owner#permits}: a missing rendezvous owner is not an authenticated
+     * unnamed primary, so that gate refuses every caller when no owner was recorded.
+     */
+    private static boolean ownsTicket(Task task, String callerOwner) {
+        return INTERNAL_NO_OWNER_CHECK.equals(callerOwner)
+                || Objects.equals(callerOwner, task.creatorOwner);
+    }
+
     /**
      * Test seam only — carries no production behaviour, and nothing in this class calls it;
      * {@link #pruneTerminalTickets} still reads {@link Task#completedNanos} directly.
@@ -1706,22 +1808,84 @@ public final class MessageService {
     }
 
     /**
-     * The question {@code workerSession} is currently paused on via {@code fleet_ask}, if any
-     * (CB-582) — {@code fleet_status} uses this to show a pending question without the caller
-     * needing the ticket. {@code null} when the session has no open async question (including a
-     * session mid a <em>blocking</em> {@code fleet_ask}, which has no {@link Task} to look up — see
-     * {@link PendingAsk}).
+     * The question {@code workerSession} is currently paused on via {@code fleet_ask}, if any —
+     * {@code fleet_status} uses this to show a pending question without the caller needing the
+     * ticket. {@code null} when the session has no open async question (including a session mid a
+     * <em>blocking</em> {@code fleet_ask}, which has no {@link Task} to look up — see
+     * {@link PendingAsk}), or when {@code callerOwner} does not own the task the question
+     * belongs to (see {@link #ownsTicket(Task, String)}).
      */
-    public PendingAsk pendingAsk(String workerSession) {
+    public PendingAsk pendingAsk(String workerSession, String callerOwner) {
         for (Task task : tasks.values()) {
             Reply q = task.question;
-            if (q != null && workerSession.equals(task.target)) {
+            if (q != null && workerSession.equals(task.target) && ownsTicket(task, callerOwner)) {
                 return new PendingAsk(task.ticket, q.text(), q.turnId());
             }
         }
         return null;
     }
 
+    /**
+     * One ticket {@code callerOwner} created, still present in {@link #tasks}, surfaced by
+     * {@link #outstanding} so a lead can carry its id into a handover file. {@link #phase} is the
+     * same value {@link #poll} would report right now, terminal phases included: a {@code DONE} or
+     * {@code FAILED} ticket stays in {@link #tasks} — and so stays reported here — until
+     * {@link #pruneTerminalTickets} evicts it.
+     */
+    public record OutstandingTicket(String ticket, Phase phase, String target) {
+    }
+
+    /**
+     * One worker session paused in {@code fleet_ask}, with the {@code turnId} that answers it,
+     * surfaced by {@link #outstanding} alongside {@link OutstandingTicket}.
+     */
+    public record OutstandingAsk(String ticket, String turnId, String workerSession) {
+    }
+
+    /** The outstanding tickets and open asks a single call to {@link #outstanding} reports. */
+    public record Outstanding(List<OutstandingTicket> tickets, List<OutstandingAsk> asks) {
+    }
+
+    /**
+     * Every ticket {@code callerOwner} created that is still in {@link #tasks} — including a
+     * finished one nobody has polled yet, since {@link #pruneTerminalTickets} discards its reply
+     * on a timer and a lead that does not carry its id forward can no longer read it after losing
+     * its session's context — plus the subset of those whose worker is paused in
+     * {@code fleet_ask}. Filtered by the same ownership rule as {@link #poll}:
+     * {@link #ownsTicket(Task, String)}.
+     */
+    public Outstanding outstanding(String callerOwner) {
+        List<OutstandingTicket> tickets = new ArrayList<>();
+        List<OutstandingAsk> asks = new ArrayList<>();
+        for (Task task : tasks.values()) {
+            if (!ownsTicket(task, callerOwner)) {
+                continue;
+            }
+            Reply question = task.question;
+            Phase phase;
+            if (task.future.isDone()) {
+                phase = terminalPhase(task.future);
+            } else if (question != null) {
+                phase = Phase.ASKING;
+                asks.add(new OutstandingAsk(task.ticket, question.turnId(), task.target));
+            } else {
+                phase = Phase.PENDING;
+            }
+            tickets.add(new OutstandingTicket(task.ticket, phase, task.target));
+        }
+        return new Outstanding(tickets, asks);
+    }
+
+    /** As {@link #poll}'s own terminal-result handling, reduced to just the {@link Phase}. */
+    private static Phase terminalPhase(CompletableFuture<Reply> future) {
+        try {
+            Reply r = future.getNow(null);
+            return r != null && r.completed() ? Phase.DONE : Phase.FAILED;
+        } catch (CompletionException | java.util.concurrent.CancellationException e) {
+            return Phase.FAILED;
+        }
+    }
+
     /** Release the async executor. */
     public void close() {
         asyncExecutor.shutdown();

fleetd/src/main/java/dev/ltms/fleet/msg/Rendezvous.java

@@ -1,5 +1,6 @@
 package dev.ltms.fleet.msg;
 
+import java.util.UUID;
 import java.util.concurrent.CompletableFuture;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.atomic.AtomicLong;
@@ -60,7 +61,7 @@ public final class Rendezvous {
     }
 
     /** A worker's open mid-turn question: the worker session it belongs to and the answer future. */
-    private record AskWaiter(String session, CompletableFuture<String> answer) {
+    private record AskWaiter(String session, CompletableFuture<String> answer, Owner owner) {
     }
 
     /**
@@ -70,11 +71,44 @@ public final class Rendezvous {
     public record AskTicket(String turnId, CompletableFuture<String> answer, boolean fresh) {
     }
 
-    private final ConcurrentHashMap<String, CompletableFuture<Resolution>> waiters = new ConcurrentHashMap<>();
+    /**
+     * The caller whose accepted delegation opened a turn — the only caller allowed to answer it.
+     * A {@code null} owner key means the unnamed primary.
+     */
+    public record Owner(String ownerKey) {
+        public static final Owner UNNAMED_PRIMARY = new Owner(null);
+
+        public static Owner of(String ownerKey) {
+            return ownerKey == null ? UNNAMED_PRIMARY : new Owner(ownerKey);
+        }
+
+        /**
+         * Whether {@code callerOwner} matches {@code owner}. A {@code null} owner means no owner was
+         * recorded, so it matches no caller. {@link #UNNAMED_PRIMARY} records the unnamed primary
+         * with an owner object whose key is {@code null}.
+         */
+        public static boolean permits(Owner owner, String callerOwner) {
+            return owner != null && java.util.Objects.equals(owner.ownerKey(), callerOwner);
+        }
+    }
+
+    /** A registered forward waiter together with the owner its delegation was opened under. */
+    private record ForwardWaiter(Owner owner, CompletableFuture<Resolution> future) {
+    }
+
+    private final ConcurrentHashMap<String, ForwardWaiter> waiters = new ConcurrentHashMap<>();
 
     /** Reverse rendezvous (CB-205): worker questions awaiting the primary's answer, keyed by {@code turnId}. */
     private final ConcurrentHashMap<String, AskWaiter> asks = new ConcurrentHashMap<>();
     private final AtomicLong askSeq = new AtomicLong();
+    /**
+     * Minted once per {@code Rendezvous} instance and folded into every {@code turnId} (see
+     * {@link #openAsk(String)}). {@link #askSeq} alone restarts at zero for every instance, so
+     * without this a {@code turnId} minted by one instance could be minted again by another and
+     * resolve to an unrelated ask with no error; this nonce makes that impossible, because an id
+     * minted by one instance can never match the id space of another.
+     */
+    private final String askBootNonce = UUID.randomUUID().toString().substring(0, 6);
     /** Per-session index of the currently-open ask, so duplicate fleet_ask calls coalesce onto one turn. */
     private final ConcurrentHashMap<String, String> openAsksBySession = new ConcurrentHashMap<>();
 
@@ -89,8 +123,17 @@ public final class Rendezvous {
      * code a double open is impossible; this is a tripwire for the day that no longer holds.
      */
     public CompletableFuture<Resolution> open(String session) {
+        return open(session, null);
+    }
+
+    /**
+     * Same as {@link #open(String)}, additionally recording {@code owner} as the caller whose
+     * delegation opened this waiter. A {@code null} owner records no owner at all — the
+     * fail-closed default {@link Owner#permits} refuses to everyone.
+     */
+    public CompletableFuture<Resolution> open(String session, Owner owner) {
         CompletableFuture<Resolution> waiter = new CompletableFuture<>();
-        CompletableFuture<Resolution> existing = waiters.putIfAbsent(session, waiter);
+        ForwardWaiter existing = waiters.putIfAbsent(session, new ForwardWaiter(owner, waiter));
         if (existing != null) {
             throw new IllegalStateException(
                     "rendezvous double-open for session " + session + " — a waiter is already registered");
@@ -105,7 +148,13 @@ public final class Rendezvous {
      * successful {@code open} after a finished turn requires this close to have happened first).
      */
     public void close(String session, CompletableFuture<Resolution> waiter) {
-        waiters.remove(session, waiter);
+        waiters.computeIfPresent(session, (s, w) -> w.future() == waiter ? null : w);
+    }
+
+    /** The owner recorded for {@code session}'s open waiter, or {@code null} if none is open. */
+    public Owner ownerOf(String session) {
+        ForwardWaiter w = waiters.get(session);
+        return w == null ? null : w.owner();
     }
 
     /** Whether a send is currently awaiting a resolution for {@code session}. */
@@ -119,7 +168,8 @@ public final class Rendezvous {
      * send (see the CB-116 note above) rather than whichever send happens to be waiting when they fire.
      */
     public CompletableFuture<Resolution> currentWaiter(String session) {
-        return waiters.get(session);
+        ForwardWaiter w = waiters.get(session);
+        return w == null ? null : w.future();
     }
 
     /**
@@ -145,9 +195,9 @@ public final class Rendezvous {
         while (true) {
             AskWaiter[] minted = { null };
             String turnId = openAsksBySession.computeIfAbsent(session, _ -> {
-                String newTurnId = session + "#" + askSeq.incrementAndGet();
+                String newTurnId = session + "#" + askBootNonce + "-" + askSeq.incrementAndGet();
                 CompletableFuture<String> answer = new CompletableFuture<>();
-                AskWaiter waiter = new AskWaiter(session, answer);
+                AskWaiter waiter = new AskWaiter(session, answer, ownerOf(session));
                 asks.put(newTurnId, waiter);
                 minted[0] = waiter;
                 return newTurnId;
@@ -183,6 +233,16 @@ public final class Rendezvous {
         return w == null ? null : w.session();
     }
 
+    /**
+     * The owner recorded for {@code turnId} when its ask turn was freshly opened — the caller
+     * whose delegation {@link #answerAsk} must match. {@code null} if {@code turnId} is unknown or
+     * lapsed, or if the ask opened with no forward waiter owner on record.
+     */
+    public Owner askOwner(String turnId) {
+        AskWaiter w = asks.get(turnId);
+        return w == null ? null : w.owner();
+    }
+
     /**
      * Resolve a worker's blocked {@code fleet_ask} with the primary's {@code answer}, unblocking it
      * to resume its turn.
@@ -245,7 +305,7 @@ public final class Rendezvous {
     }
 
     private boolean complete(String session, Resolution resolution) {
-        CompletableFuture<Resolution> waiter = waiters.get(session);
-        return waiter != null && waiter.complete(resolution);
+        ForwardWaiter waiter = waiters.get(session);
+        return waiter != null && waiter.future().complete(resolution);
     }
 }

fleetd/src/main/java/dev/ltms/fleet/msg/ReplyPushLoop.java

@@ -3,6 +3,7 @@ package dev.ltms.fleet.msg;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
 import dev.ltms.fleet.herdr.HerdrException;
+import dev.ltms.fleet.herdr.PromptBox;
 import dev.ltms.fleet.mcp.PrimaryRegistry;
 import dev.ltms.fleet.metrics.FleetMetrics;
 import dev.ltms.fleet.metrics.Metrics;
@@ -19,6 +20,7 @@ import java.util.Set;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.ScheduledExecutorService;
 import java.util.concurrent.TimeUnit;
+import java.util.function.BiPredicate;
 import java.util.stream.Collectors;
 
 /**
@@ -50,6 +52,11 @@ import java.util.stream.Collectors;
  * exhausting its cap does not stop nudges about the others (post-CB-590 regression fix; see
  * {@link #decide}) — whichever the durable inbox / pending set doesn't already answer via
  * {@code STOP}.
+ *
+ * <p>A lead that is injectable is nudged only when its prompt box is also empty
+ * ({@link PromptBox}): the delivery pastes and submits in one call, so a nudge into a box holding
+ * the operator's half-typed line would submit that line too. A nudge held for that reason waits for
+ * the next tick like any other, and the pending work is re-read then.
  */
 public final class ReplyPushLoop {
 
@@ -81,11 +88,19 @@ public final class ReplyPushLoop {
 
     private final PrimaryRegistry primaryRegistry;
     private final AgentControl agents;
+    private final PromptBox promptBox;
     private final ReplyInbox inbox;
     private final ScheduledExecutorService scheduler;
     private final int maxReminders;
     private final long backoffMs;
     private final Metrics metrics; // CB-512: nullable — no registry in unit tests
+    /**
+     * Whether the resolved lead may still poll a given ticket, consulted before a nudge ever names
+     * it. {@code fleet_send(wait:false)} is reachable by a collaborator or an observer, neither of
+     * which this loop's own identity ladder checks, so left unguarded a nudge could instruct a pane
+     * to run a call its authorization would refuse.
+     */
+    private final BiPredicate<String, String> ticketPollAuthorized;
 
     /**
      * Worker targets with a reply queued, keyed by target. Each entry carries its own nudge
@@ -123,13 +138,29 @@ public final class ReplyPushLoop {
     public ReplyPushLoop(PrimaryRegistry primaryRegistry, AgentControl agents, ReplyInbox inbox,
                          ScheduledExecutorService scheduler,
                          int maxReminders, long backoffMs, Metrics metrics) {
+        this(primaryRegistry, agents, inbox, scheduler, maxReminders, backoffMs, metrics,
+                (lead, ticket) -> true);
+    }
+
+    /**
+     * As above, with {@code ticketPollAuthorized} wired to the real authorization check. The two
+     * constructors above pass a predicate that always answers yes, matching this loop's behaviour
+     * before the check existed — the real production wiring supplies the one backed by the
+     * authorization table.
+     */
+    public ReplyPushLoop(PrimaryRegistry primaryRegistry, AgentControl agents, ReplyInbox inbox,
+                         ScheduledExecutorService scheduler,
+                         int maxReminders, long backoffMs, Metrics metrics,
+                         BiPredicate<String, String> ticketPollAuthorized) {
         this.primaryRegistry = primaryRegistry;
         this.agents = agents;
+        this.promptBox = new PromptBox(agents);
         this.inbox = inbox;
         this.scheduler = scheduler;
         this.maxReminders = maxReminders;
         this.backoffMs = backoffMs;
         this.metrics = metrics;
+        this.ticketPollAuthorized = ticketPollAuthorized == null ? (lead, ticket) -> true : ticketPollAuthorized;
     }
 
     /**
@@ -186,9 +217,27 @@ public final class ReplyPushLoop {
     private record PendingTicket(String ticket, String lead, boolean failed, int nudgeCount) {
     }
 
-    /** Tickets still pending for {@code lead}, snapshotted fresh for one tick. */
+    /**
+     * Tickets still pending for {@code lead}, snapshotted fresh for one tick — excluding any ticket
+     * {@link #ticketPollAuthorized} says {@code lead} may not poll. Filtered here, at the one source
+     * every other lookup in this class reads from, so an unauthorized ticket is invisible to
+     * {@link #decide}, {@link #injectNudge}, and {@link #bumpNudgeCounts} alike: it is never named
+     * in a nudge, never spends its nudge budget, and never keeps a schedule alive on its own.
+     */
     private List<PendingTicket> pendingTicketsFor(String lead) {
-        return pendingTickets.values().stream().filter(t -> lead.equals(t.lead())).toList();
+        return pendingTickets.values().stream()
+                .filter(t -> lead.equals(t.lead()))
+                .filter(t -> authorizedToPoll(lead, t.ticket()))
+                .toList();
+    }
+
+    private boolean authorizedToPoll(String lead, String ticket) {
+        if (ticketPollAuthorized.test(lead, ticket)) {
+            return true;
+        }
+        log.debug("push: ticket {} is pending for {} but it may not poll it, suppressing the nudge",
+                ticket, lead);
+        return false;
     }
 
     /** Ticket ids still pending for {@code lead} — a plain snapshot for race comparison. */
@@ -398,11 +447,15 @@ public final class ReplyPushLoop {
             log.debug("push: status check failed for lead {}, will retry", lead, e);
             return Action.WAIT_BUSY;
         }
-        if (status.injectable()) {
-            return Action.INJECT;
+        if (!status.injectable()) {
+            log.debug("push: lead {} is {} (not injectable), waiting", lead, status);
+            return Action.WAIT_BUSY;
         }
-        log.debug("push: lead {} is {} (not injectable), waiting", lead, status);
-        return Action.WAIT_BUSY;
+        if (!promptBox.clearToSubmit(lead)) {
+            log.debug("push: lead {} has unsubmitted text in its prompt box, waiting", lead);
+            return Action.WAIT_BUSY;
+        }
+        return Action.INJECT;
     }
 
     /**
@@ -439,6 +492,15 @@ public final class ReplyPushLoop {
      * — timeout, transport error, a decode error — is treated as still live and the binding is left
      * alone, because guessing wrong here is unrecoverable while guessing "live" merely costs one more
      * retry on the next tick, which {@link #decide} already tolerates.
+     *
+     * <p><strong>The fallback is probed too.</strong> {@code PrimaryRegistry.nudgeTargetFor} already
+     * resolves a named delegator to its current terminal before this method ever sees it, which
+     * keeps a rolled lead's per-target binding live. What that resolution cannot fix is a caller
+     * that was never recorded with a name at all — an unnamed primary, or a lead whose tab the
+     * scanner cannot currently see — where the fallback it returns is still the raw terminal last
+     * learned from call traffic. This method returns that fallback only after the same liveness
+     * check, and gives up for this tick (an empty result, exactly like "no lead known at all") rather
+     * than hand a caller a second stale address un-probed.
      */
     private Optional<String> resolveLiveLead(String target) {
         Optional<String> lead = primaryRegistry.nudgeTargetFor(target);
@@ -448,7 +510,12 @@ public final class ReplyPushLoop {
         log.debug("push: lead {} delegated to for {} is no longer live, forgetting the stale binding "
                 + "and falling back", lead.get(), target);
         primaryRegistry.forgetDelegation(target);
-        return primaryRegistry.nudgeTargetFor(target);
+        Optional<String> fallback = primaryRegistry.nudgeTargetFor(target);
+        if (fallback.isEmpty() || isLive(fallback.get())) {
+            return fallback;
+        }
+        log.debug("push: fallback lead {} for {} is also not live, skipping this tick", fallback.get(), target);
+        return Optional.empty();
     }
 
     /**
@@ -753,7 +820,7 @@ public final class ReplyPushLoop {
         String nudge = formatNudge(replyTargets, tickets, questions, incidents, unmappedTargets);
         try {
             agents.send(lead, nudge);
-            log.debug("push: nudge sent to lead {} (reply {}/{}, ticket {}/{}, question {}/{}; "
+            log.debug("push: nudge sent to {} (reply {}/{}, ticket {}/{}, question {}/{}; "
                             + "{} reply target(s), {} ticket(s), {} question(s))",
                     lead, replyReminderCount + 1, maxReminders, ticketReminderCount + 1, maxReminders,
                     questionReminderCount + 1, maxReminders,
@@ -770,7 +837,7 @@ public final class ReplyPushLoop {
                 }
             }
         } catch (RuntimeException e) {
-            log.warn("push: failed to nudge lead {} (reply {}/{}, ticket {}/{}, question {}/{}): {}",
+            log.warn("push: failed to nudge {} (reply {}/{}, ticket {}/{}, question {}/{}): {}",
                     lead, replyReminderCount + 1, maxReminders, ticketReminderCount + 1, maxReminders,
                     questionReminderCount + 1, maxReminders, e.toString());
         }

fleetd/src/main/java/dev/ltms/fleet/peer/MemberRole.java

@@ -1,6 +1,8 @@
 package dev.ltms.fleet.peer;
 
 import java.util.Locale;
+import java.util.stream.Collectors;
+import java.util.stream.Stream;
 
 /**
  * What a member is <em>for</em> — the contract it runs under.
@@ -100,6 +102,11 @@ public enum MemberRole {
         return null;
     }
 
+    /** The wire name of every role, joined with {@code ", "} in declaration order. */
+    public static String wireNames() {
+        return Stream.of(values()).map(MemberRole::wireName).collect(Collectors.joining(", "));
+    }
+
     /**
      * Parse a config/wire spelling, case-insensitively.
      *
@@ -118,14 +125,7 @@ public enum MemberRole {
                 }
             }
         }
-        StringBuilder valid = new StringBuilder();
-        for (MemberRole r : values()) {
-            if (!valid.isEmpty()) {
-                valid.append(", ");
-            }
-            valid.append(r.wireName());
-        }
         throw new IllegalArgumentException(
-                "unknown member role '" + s + "'; valid roles are: " + valid);
+                "unknown member role '" + s + "'; valid roles are: " + wireNames());
     }
 }

fleetd/src/main/java/dev/ltms/fleet/rest/FleetApp.java

@@ -13,6 +13,7 @@ import dev.ltms.fleet.mcp.FleetMcp;
 import dev.ltms.fleet.herdr.Agent;
 import dev.ltms.fleet.herdr.HerdrClient;
 import dev.ltms.fleet.herdr.HerdrException;
+import dev.ltms.fleet.herdr.PaneLocator;
 import dev.ltms.fleet.inject.MemberPresence;
 import dev.ltms.fleet.member.MemberCredentialPolicyView;
 import dev.ltms.fleet.peer.PeerUnreachableException;
@@ -49,22 +50,52 @@ import java.util.stream.Collectors;
  */
 public final class FleetApp {
 
-    /** The authorization action the matching route handler hands to {@link #allow}. */
+    /**
+     * The authorization action the matching route handler hands to {@link #allow}, for a route
+     * whose action does not depend on the request body.
+     */
     static Authz.Action routeAction(String route) {
+        return routeAction(route, null);
+    }
+
+    /**
+     * As above, plus the one route whose action depends on the body: {@code POST
+     * /sessions/{id}/message} carries a {@code turnId} (the answer-a-blocked-worker shape) or not
+     * (a plain delivery), mirroring {@code FleetMcp#sendAction}'s split of the same two call
+     * shapes over MCP. {@code turnId} is ignored by every other route.
+     *
+     * @param turnId the request body's {@code turnId}, or {@code null}/blank when absent or not
+     *               applicable to this route
+     */
+    static Authz.Action routeAction(String route, String turnId) {
         return switch (route) {
             case "GET /metrics" -> Authz.Action.METRICS;
             case "POST /members" -> Authz.Action.SPAWN;
             case "DELETE /members/{paneId}" -> Authz.Action.STOP;
-            case "POST /sessions/{id}/message" -> Authz.Action.SEND;
+            case "POST /sessions/{id}/message" -> turnId == null || turnId.isBlank()
+                    ? Authz.Action.SEND : Authz.Action.ANSWER;
             case "POST /sessions/{id}/reply" -> Authz.Action.REPLY;
             case "GET /sessions/{id}/replies" -> Authz.Action.DRAIN;
             case "POST /sessions/{id}/ask" -> Authz.Action.ASK;
             case "GET /sessions", "GET /agents", "GET /members", "GET /profiles",
-                    "GET /member-credentials", "GET /sessions/{id}/status", "GET /tasks/{ticket}" -> Authz.Action.READ;
+                    "GET /member-credentials" -> Authz.Action.READ;
+            case "GET /sessions/{id}/status", "GET /tasks/{ticket}" -> Authz.Action.TASK_READ;
             default -> throw new IllegalArgumentException("route has no authorization gate: " + route);
         };
     }
 
+    /**
+     * The second gate for {@code POST /sessions/{id}/message}: checked only when {@code turnId}
+     * is present and non-blank, against {@link Authz.Action#ANSWER}. A request with no {@code
+     * turnId} passes this gate unconditionally, without consulting {@code permit} at all, having
+     * already cleared the coarse {@link Authz.Action#SEND} grant checked ahead of it.
+     *
+     * @param permit reports whether the caller holds the named grant
+     */
+    static boolean answerGatePasses(String turnId, Predicate<Authz.Action> permit) {
+        return turnId == null || turnId.isBlank() || permit.test(Authz.Action.ANSWER);
+    }
+
     /** Default blocking window for a message; kept under typical HTTP idle timeouts. */
     private static final long DEFAULT_MESSAGE_TIMEOUT_MS = 25_000;
     private static final long MAX_MESSAGE_TIMEOUT_MS = 120_000;
@@ -236,6 +267,44 @@ public final class FleetApp {
         return app;
     }
 
+    /**
+     * The authorization decision behind {@link #allow}, taking the caller directly rather than
+     * pulling it from a servlet {@link Context} — unit-testable without fabricating a live
+     * request, the same reason {@code FleetMcp#denyFor} is split from {@code FleetMcp#deny}.
+     *
+     * @param knownLeadOrCollaborator the classifier a collaborator's {@code SEND} is checked
+     *                                against; pass {@link #auth}'s own {@code
+     *                                knownLeadOrCollaborator()} to exercise the real production
+     *                                gate, as {@link #allow} does
+     */
+    static boolean permitsFor(Principal caller, Authz.Action action, String target,
+                              Predicate<String> knownLeadOrCollaborator) {
+        return Authz.permits(caller, action, target, knownLeadOrCollaborator);
+    }
+
+    /**
+     * As {@link #permitsFor(Principal, Authz.Action, String, Predicate)}, also threading the
+     * classifier an observer's {@code SEND} is checked against; pass {@link #auth}'s own
+     * {@code sendableObserverTarget()} to exercise the real production gate, as {@link #allow}
+     * does.
+     */
+    static boolean permitsFor(Principal caller, Authz.Action action, String target,
+                              Predicate<String> knownLeadOrCollaborator,
+                              Predicate<String> knownObserverTarget) {
+        return Authz.permits(caller, action, target, knownLeadOrCollaborator, knownObserverTarget);
+    }
+
+    /**
+     * As {@link #permitsFor(Principal, Authz.Action, String, Predicate, Predicate)}, also
+     * threading the classifier {@code TASK_READ}'s ticket-ownership grant is checked against.
+     */
+    static boolean permitsFor(Principal caller, Authz.Action action, String target,
+                              Predicate<String> knownLeadOrCollaborator,
+                              Predicate<String> knownObserverTarget,
+                              Predicate<String> callerOwnsTicket) {
+        return Authz.permits(caller, action, target, knownLeadOrCollaborator, knownObserverTarget, callerOwnsTicket);
+    }
+
     /**
      * Gate a handler on the CB-505 authorization table. Returns {@code true} when the request may
      * proceed; otherwise writes the error response and returns {@code false}.
@@ -245,12 +314,22 @@ public final class FleetApp {
      * yours" (a worker reaching for another worker's session, or for orchestration).
      */
     private boolean allow(Context ctx, Authz.Action action, String target) {
+        return allow(ctx, action, target, Authz.NO_CALLER_OWNS_TICKET);
+    }
+
+    /**
+     * As {@link #allow(Context, Authz.Action, String)}, also threading the classifier
+     * {@code TASK_READ}'s ticket-ownership grant is checked against.
+     */
+    private boolean allow(Context ctx, Authz.Action action, String target, Predicate<String> callerOwnsTicket) {
         if (auth == null) {
             return true; // legacy: authorization not enforced
         }
         Principal caller = ctx.attribute(CALLER);
-        if (Authz.permits(caller, action, target)) {
-            if (action != Authz.Action.READ && action != Authz.Action.METRICS) {
+        if (permitsFor(caller, action, target, auth.knownLeadOrCollaborator(), auth.sendableObserverTarget(),
+                callerOwnsTicket)) {
+            if (action != Authz.Action.READ && action != Authz.Action.METRICS
+                    && action != Authz.Action.TASK_READ) {
                 AuditLog.allowed(caller, action, target); // reads would drown the trail
             }
             return true;
@@ -390,14 +469,27 @@ public final class FleetApp {
         }
     }
 
+    /**
+     * Tab id → its herdr display label, or an empty map on a {@code workspace.list}/{@code
+     * tab.list} failure — a missing label must not cost the agent roster.
+     */
+    private Map<String, String> tabLabelsOrEmpty() {
+        try {
+            return new PaneLocator(herdr, memberHerdr).tabLabelsByTabId();
+        } catch (HerdrException e) {
+            return Map.of();
+        }
+    }
+
     /** Discovery: every agent herdr tracks, keyed by its Claude session UUID. */
     private void agents(Context ctx) {
         if (!allow(ctx, routeAction("GET /agents"), null)) {
             return;
         }
+        final Map<String, String> tabLabels = tabLabelsOrEmpty();
         try {
             ctx.status(200).json(Map.of("agents",
-                    workers.list().stream().map(Agent.class::cast).map(FleetApp::view).toList()));
+                    workers.list().stream().map(Agent.class::cast).map(a -> view(a, tabLabels)).toList()));
         } catch (HerdrException e) {
             // fleetd #297: workers.list() reaches herdr — a transport failure must land in the same
             // {error, detail} envelope every other failure path here uses, not escape as a bare
@@ -603,47 +695,64 @@ public final class FleetApp {
      * status-gated injector and block until the worker returns a structured {@code fleet_reply}.
      * Times out with a typed 202 (working / queued / busy) rather than an error — the message may
      * still land.
+     *
+     * <p>Two call shapes share this route, exactly as {@code fleet_send} does over MCP (see
+     * {@code FleetMcp#sendAction}): a plain delivery to {@code id}, and -- when the body carries
+     * {@code turnId} -- resolving a worker's blocked question. The coarse {@link
+     * Authz.Action#SEND} grant is checked first, before the body is read at all; only once that
+     * passes is the body parsed, and a present {@code turnId} is then checked again against
+     * {@link Authz.Action#ANSWER}. A body that fails to parse is rejected with 400 and reaches
+     * neither {@code messages.answer} nor {@code messages.send}.
      */
     private void sendMessage(Context ctx) {
         String id = ctx.pathParam("id");
         if (!allow(ctx, routeAction("POST /sessions/{id}/message"), id)) {
             return;
         }
-        String content;
-        String turnId;
-        long timeout;
-        boolean wait;
+        Principal caller = ctx.attribute(CALLER);
+        String callerOwner = caller == null ? null : caller.ownerKey();
+        JsonNode body;
         try {
-            JsonNode body = mapper.readTree(ctx.body());
-            content = body.path("content").asText("");
-            turnId = body.path("turnId").asText(null);
-            timeout = body.path("timeoutMs").asLong(DEFAULT_MESSAGE_TIMEOUT_MS);
-            wait = body.path("wait").asBoolean(true); // default: block for the reply (CB-104)
+            body = mapper.readTree(ctx.body());
         } catch (Exception e) {
+            body = null;
+        }
+        if (body == null) {
             ctx.status(400).json(Map.of("error", "bad_request", "detail", "body must be JSON"));
             return;
         }
+        String turnId = body.path("turnId").asText(null);
+        if (!answerGatePasses(turnId, action -> allow(ctx, action, id))) {
+            return;
+        }
+        String content = body.path("content").asText("");
+        long timeout = body.path("timeoutMs").asLong(DEFAULT_MESSAGE_TIMEOUT_MS);
+        boolean wait = body.path("wait").asBoolean(true); // default: block for the reply (CB-104)
         if (content.isBlank()) {
             ctx.status(400).json(Map.of("error", "bad_request", "detail", "content is required"));
             return;
         }
+        // An observer's SEND reaches a pane that cannot otherwise distinguish this from a human
+        // paste (see FleetMcp#attributeIfObserver, the same rule on the MCP entry path); every
+        // other caller's content passes through unchanged.
+        content = FleetMcp.attributeIfObserver(caller, content);
         timeout = Math.clamp(timeout, 1, MAX_MESSAGE_TIMEOUT_MS);
 
         // Answering a worker's fleet_ask (CB-205): always blocks, and derives the worker from turnId.
         if (turnId != null && !turnId.isBlank()) {
-            writeReply(ctx, id, messages.answer(turnId, content, timeout), timeout);
+            writeReply(ctx, id, messages.answer(turnId, content, timeout, callerOwner), timeout);
             return;
         }
 
         if (!wait) {
             // Fire-and-poll (CB-107): return a ticket immediately; the caller polls GET /tasks/{ticket}.
-            String ticket = messages.sendAsync(id, content);
+            String ticket = messages.sendAsync(id, content, null, caller);
             ctx.status(202).json(Map.of("sessionId", id, "ticket", ticket, "status", "accepted"));
             return;
         }
 
         try {
-            writeReply(ctx, id, messages.send(id, content, timeout), timeout);
+            writeReply(ctx, id, messages.send(id, content, timeout, callerOwner), timeout);
         } catch (HerdrException e) {
             herdrError(ctx, e);
         }
@@ -662,6 +771,10 @@ public final class FleetApp {
             case STALE_TURN -> ctx.status(409).json(Map.of(
                     "sessionId", id, "error", "stale_turn",
                     "detail", "that question is no longer open (timed out or already answered)"));
+            case NOT_TURN_OWNER -> ctx.status(403).json(Map.of(
+                    "sessionId", id, "error", "not_turn_owner",
+                    "detail", "this turn belongs to a different delegation — only the caller that "
+                            + "opened it may answer it"));
             case REPLIED, COMPLETED_UNREPLIED -> {
                 // replySource distinguishes a structured fleet_reply from the CB-106 completion
                 // fallback (a scrape of the worker's transcript when it finished without replying).
@@ -676,9 +789,10 @@ public final class FleetApp {
                     // silent fall-through. That is exactly the bug this ticket exists to fix:
                     // `default -> "done"` used to sit here and would have told a REST caller the
                     // delegation completed for TIMED_OUT_UNCONFIRMED, the one outcome where delivery
-                    // is unknown. REPLIED, COMPLETED_UNREPLIED, QUESTION and STALE_TURN can never
-                    // actually reach this inner switch — the outer switch above always dispatches
-                    // them first — but they still need an arm to keep this switch exhaustive.
+                    // is unknown. REPLIED, COMPLETED_UNREPLIED, QUESTION, STALE_TURN and
+                    // NOT_TURN_OWNER can never actually reach this inner switch — the outer switch
+                    // above always dispatches them first — but they still need an arm to keep this
+                    // switch exhaustive.
                     "status", switch (reply.outcome()) {
                         case TIMED_OUT_WORKING -> "working";
                         case TIMED_OUT_QUEUED -> "queued";
@@ -688,7 +802,7 @@ public final class FleetApp {
                         case BUSY -> "busy";
                         case WORKER_FAILED -> "failed";
                         case BACKEND_EXHAUSTED -> "backend_exhausted";
-                        case REPLIED, COMPLETED_UNREPLIED, QUESTION, STALE_TURN -> "done"; // unreachable
+                        case REPLIED, COMPLETED_UNREPLIED, QUESTION, STALE_TURN, NOT_TURN_OWNER -> "done"; // unreachable
                     },
                     "detail", reply.outcome() == MessageService.Outcome.TIMED_OUT_UNCONFIRMED
                             ? "no reply within " + timeout + "ms; delivery is unconfirmed — the "
@@ -817,10 +931,11 @@ public final class FleetApp {
             body.put("sessionId", id);
             body.put("status", messages.status(id).name().toLowerCase());
             body.put("ready", deliverable.test(id));
-            // CB-582: a worker paused mid-turn in an async fleet_ask is otherwise invisible to a
-            // status poll — surface the open question and how to answer it, same as fleet_poll's
-            // Phase.ASKING view.
-            MessageService.PendingAsk ask = messages.pendingAsk(id);
+            // A worker paused mid-turn in an async fleet_ask is otherwise invisible to a status
+            // poll — surface the open question and how to answer it, same as fleet_poll's
+            // Phase.ASKING view, but only to the caller whose owner key created that delegation.
+            Principal caller = ctx.attribute(CALLER);
+            MessageService.PendingAsk ask = messages.pendingAsk(id, caller == null ? null : caller.ownerKey());
             if (ask != null) {
                 body.put("question", ask.question());
                 body.put("turnId", ask.turnId());
@@ -834,10 +949,14 @@ public final class FleetApp {
 
     /** Poll an async (wait:false) delegation by ticket. 404 for an unknown/expired ticket. */
     private void taskStatus(Context ctx) {
-        if (!allow(ctx, routeAction("GET /tasks/{ticket}"), null)) {
+        String ticket = ctx.pathParam("ticket");
+        Principal caller = ctx.attribute(CALLER);
+        String ownerKey = caller == null ? null : caller.ownerKey();
+        if (!allow(ctx, routeAction("GET /tasks/{ticket}"), ticket,
+                t -> t != null && messages.isTicketOwnedBy(t, ownerKey))) {
             return;
         }
-        MessageService.TaskView v = messages.poll(ctx.pathParam("ticket"));
+        MessageService.TaskView v = messages.poll(ticket, ownerKey);
         if (v == null) {
             ctx.status(404).json(Map.of("error", "unknown_ticket", "detail", "no such task (or it has expired)"));
             return;
@@ -869,13 +988,19 @@ public final class FleetApp {
         }
     }
 
-    /** Stable JSON projection of an agent (null-safe for the start-time shape). */
-    private static Map<String, Object> view(Agent a) {
+    /**
+     * Stable JSON projection of an agent (null-safe for the start-time shape).
+     *
+     * @param tabLabels tab id → its herdr display label; a tab absent from this map, or carrying
+     *                   a {@code null} label itself, projects as a {@code null} "label"
+     */
+    private static Map<String, Object> view(Agent a, Map<String, String> tabLabels) {
         Map<String, Object> m = new LinkedHashMap<>();
         m.put("terminalId", a.terminalId());
         m.put("paneId", a.paneId());
         m.put("workspaceId", a.workspaceId());
         m.put("tabId", a.tabId());
+        m.put("label", tabLabels.get(a.tabId()));
         m.put("sessionId", a.sessionId());
         m.put("agentType", a.agentType());
         m.put("status", a.status().name().toLowerCase());

fleetd/src/main/java/dev/ltms/fleet/session/GitWorktrees.java

@@ -442,38 +442,6 @@ public final class GitWorktrees implements Worktrees {
                 ENVIRONMENT_CREDENTIAL_HELPER);
     }
 
-    /**
-     * {@link #configureEnvironmentCredentialHelper} only ever fires for an HTTPS origin — Git never
-     * consults a {@code credential.helper} for an SSH transport. This repo's own origin is
-     * {@code ssh://git@git.ltms.dev:2224/fleet/fleetd.git}, so a member sitting on that origin never
-     * reaches the helper and the repo-scoped {@code WORKER_GITEA_TOKEN} is simply not used.
-     *
-     * <p>An earlier version of this javadoc justified the rewrite by claiming a member <em>cannot</em>
-     * push once {@code memberCredentials.policy: allow-list} blocks {@code SSH_AUTH_SOCK}, because
-     * "there is no private key file on this host, only an ssh-agent socket". That premise is false
-     * (fleetd #184): {@code ssh -G} resolves a readable, passphrase-free {@code IdentityFile} outside
-     * {@code ~/.ssh}, and a member — same OS user — pushes over SSH with the socket blanked. The
-     * rewrite is still worth having, but for the reason below rather than that one: it routes the
-     * member through its own scoped token instead of the operator's ssh identity, which is what makes
-     * a member's pushes attributable and revocable.
-     *
-     * <p>The fix is a <em>worktree-scoped</em> URL rewrite: {@code url.<https-base>.insteadOf
-     * <ssh-base>}, set with {@code --worktree} so it lands only in
-     * {@code <worktree>/.git/worktrees/<name>/config.worktree} (enabled by
-     * {@code extensions.worktreeConfig}, already turned on above) and never touches the shared
-     * repo-level config the primary checkout also reads. {@code insteadOf} — not
-     * {@code pushInsteadOf} — because a member may also need to fetch or rebase, and both should go
-     * through the member's own token for the same reason.
-     *
-     * <p>The host (and, for the rewrite's SSH-side match, the port) come from parsing the origin
-     * itself — never a hardcoded forge host, which is exactly what #177 removed. An origin that is
-     * already {@code https://} is left alone; the credential helper already covers it. An origin
-     * that is neither {@code ssh://} nor {@code https://} — including the scp-like shorthand
-     * ({@code git@host:path}, no scheme) — is left untouched deliberately: that shorthand's
-     * {@code host:path} split is defined by the user's ssh_config aliases, not by URI syntax, so
-     * guessing at it risks rewriting to the wrong place. A repo provisioned from that form keeps
-     * today's (broken, if the policy blocks the agent) SSH-only behaviour rather than a wrong rewrite.
-     */
     /**
      * Blank the user-info of a remote URL before it reaches a log. A remote URL is not obviously a
      * credential channel, which is exactly why one has leaked here three times ({@code git remote -v}
@@ -485,6 +453,30 @@ public final class GitWorktrees implements Worktrees {
         return url == null ? null : url.replaceAll("://[^@/]*@", "://<redacted>@");
     }
 
+    /**
+     * {@link #configureEnvironmentCredentialHelper} only fires for an HTTPS origin — Git never
+     * consults a {@code credential.helper} for an SSH transport. This repo's own origin is
+     * {@code ssh://git@git.ltms.dev:2224/fleet/fleetd.git}, so a member on that origin never
+     * reaches the helper, and the repo-scoped token goes unused without a separate rewrite.
+     *
+     * <p>This method routes the member through its own scoped token instead of the operator's ssh
+     * identity, which is what makes a member's pushes attributable and revocable.
+     *
+     * <p>The fix is a <em>worktree-scoped</em> URL rewrite: {@code url.<https-base>.insteadOf
+     * <ssh-base>}, set with {@code --worktree} so it lands only in
+     * {@code <worktree>/.git/worktrees/<name>/config.worktree} and never touches the shared
+     * repo-level config the primary checkout also reads. {@code insteadOf} — not
+     * {@code pushInsteadOf} — because a member may also need to fetch or rebase through its own
+     * token.
+     *
+     * <p>The host (and, for the rewrite's SSH-side match, the port) come from parsing the origin
+     * itself, never a hardcoded forge host. An origin already {@code https://} is left alone; the
+     * credential helper already covers it. An origin that is neither {@code ssh://} nor
+     * {@code https://} — including the scp-like shorthand ({@code git@host:path}, no scheme) — is
+     * left untouched: that shorthand's {@code host:path} split is defined by the user's ssh_config
+     * aliases, not by URI syntax, so guessing at it risks rewriting to the wrong place, and that
+     * origin keeps SSH-only push behaviour instead.
+     */
     private void configureHttpsUrlRewriteForSshOrigin(String repoRoot, String worktreePath) {
         if (exitCode("git", "-C", repoRoot, "config", "--get", "remote.origin.url") != 0) {
             return;

fleetd/src/main/java/dev/ltms/fleet/session/SessionManager.java

@@ -26,6 +26,7 @@ import java.util.concurrent.atomic.AtomicBoolean;
 import java.util.concurrent.atomic.AtomicLong;
 import java.util.function.Consumer;
 import java.util.function.LongSupplier;
+import java.util.function.Supplier;
 
 /**
  * Authoritative in-daemon registry of the worker sessions this {@code fleetd} process spawned.
@@ -57,6 +58,18 @@ public final class SessionManager implements TurnListener {
      * Populated on every spawn path, removed on {@link #release}.
      */
     private final ConcurrentHashMap<String /*paneId*/, PeerHandle> handles = new ConcurrentHashMap<>();
+    /**
+     * fleetd #702: a pane mid-teardown, keyed by paneId, held from just before its registry entry
+     * is removed until {@link #releaseRemoved} finishes. {@link #spawnedMemberRole} consults this
+     * alongside the registry, so a caller resolving the pane's terminal during that window still
+     * sees a live member and never falls through to a tab map.
+     *
+     * <p>Depth-counted rather than a plain set: two threads can be tearing down the same pane at
+     * once (the CAS in {@link #releaseIfCurrent} exists for exactly that race), and with a set the
+     * loser's {@code finally} would unmark the pane while the winner is still mid-teardown,
+     * reopening the window this exists to close.
+     */
+    private final ConcurrentHashMap<String /*paneId*/, Releasing> releasing = new ConcurrentHashMap<>();
     private final MemberPresence presence;
     private final SecureRandom nonceRandom = new SecureRandom();
     private final AtomicLong nonceSeq = new AtomicLong();
@@ -242,6 +255,10 @@ public final class SessionManager implements TurnListener {
                         handle.id(), handle.terminalId(), resolvedProfile, actualRole, cwd, ownerTerminal, now, now, 0,
                         MemberSession.State.SPAWNING, null, null, handle.charterReceipt(), handle.agentSessionId());
                 registry.put(handle.id(), session);
+                // A presence contact that already arrived for this terminal found no registry
+                // entry to transition and gave up silently. Retry it now that one exists; remove
+                // this call and such a session stays in SPAWNING even though it is present.
+                reconcilePresence(handle.terminalId());
                 handles.put(handle.id(), handle);
                 log.debug("acquired session id={} terminal={} profile={} owner={}",
                         handle.id(), handle.terminalId(), session.profile(), session.ownerTerminal());
@@ -303,9 +320,11 @@ public final class SessionManager implements TurnListener {
      * is a logged path an operator can reclaim, the cost of a deleted one is unrecoverable work.
      */
     private MemberSession release(String paneId, ReleaseCause cause) {
-        MemberSession removed = registry.remove(paneId);
-        releaseRemoved(paneId, removed, handles.remove(paneId), cause);
-        return removed;
+        return releaseWindow(paneId, registry.get(paneId), () -> {
+            MemberSession removed = registry.remove(paneId);
+            releaseRemoved(paneId, removed, handles.remove(paneId), cause);
+            return removed;
+        });
     }
 
     /**
@@ -314,16 +333,85 @@ public final class SessionManager implements TurnListener {
      * DONE record from stopping a worker that delivery has made BUSY.
      */
     private boolean releaseIfCurrent(MemberSession expected, ReleaseCause cause) {
-        if (!registry.remove(expected.paneId(), expected)) {
-            // A lifecycle transition replaced the record between the caller's check and this remove.
-            // Log it: this race is by definition unobservable otherwise, and a reaper that silently
-            // declines to reap is the hardest kind of behaviour to diagnose after the fact.
-            log.debug("skipping reap of pane={}: its registry record changed after the idle check "
-                    + "(most likely a delivery made it BUSY)", expected.paneId());
-            return false;
+        return releaseWindow(expected.paneId(), expected, () -> {
+            if (!registry.remove(expected.paneId(), expected)) {
+                // A lifecycle transition replaced the record between the caller's check and this
+                // remove. Log it: this race is by definition unobservable otherwise, and a reaper
+                // that silently declines to reap is the hardest kind of behaviour to diagnose
+                // after the fact.
+                log.debug("skipping reap of pane={}: its registry record changed after the idle "
+                        + "check (most likely a delivery made it BUSY)", expected.paneId());
+                return false;
+            }
+            releaseRemoved(expected.paneId(), expected, handles.remove(expected.paneId()), cause);
+            return true;
+        });
+    }
+
+    /**
+     * fleetd #702: mark {@code paneId} as mid-teardown — using {@code known}'s terminal/role when
+     * it is available — for the whole of {@code teardown}, which removes the registry entry and
+     * then runs {@link #releaseRemoved}. Shared by both registry-removal sites ({@link #release}'s
+     * unconditional remove and {@link #releaseIfCurrent}'s CAS remove) so neither can leave the
+     * other's window unmarked.
+     *
+     * <p>The mark is written before {@code teardown} runs — so it covers the removal itself, not
+     * only what comes after it — and cleared in a {@code finally}, so an unchecked throw out of
+     * {@code teardown} (including one from {@link PeerLauncher#stop}, which declares nothing) can
+     * never leave the pane marked for the rest of the daemon's life.
+     */
+    private <T> T releaseWindow(String paneId, MemberSession known, Supplier<T> teardown) {
+        releasing.compute(paneId, (_, prior) -> Releasing.enter(prior, known));
+        try {
+            return teardown.get();
+        } finally {
+            releasing.compute(paneId, (_, prior) -> prior == null ? null : prior.leave());
         }
-        releaseRemoved(expected.paneId(), expected, handles.remove(expected.paneId()), cause);
-        return true;
+    }
+
+    /**
+     * Depth count plus the terminal/role a mid-teardown pane belongs to, for
+     * {@link #spawnedMemberRole}. The terminal/role come from whichever call into
+     * {@link #releaseWindow} first knew them: a call that finds the registry entry already gone
+     * passes a {@code null} session, and must not blank out what the first call recorded.
+     */
+    record Releasing(int depth, String terminalId, MemberRole role) {
+        static Releasing enter(Releasing prior, MemberSession known) {
+            int depth = (prior == null ? 0 : prior.depth()) + 1;
+            String terminalId = known != null ? known.terminalId() : prior == null ? null : prior.terminalId();
+            MemberRole role = known != null ? known.role() : prior == null ? null : prior.role();
+            return new Releasing(depth, terminalId, role);
+        }
+
+        Releasing leave() {
+            return depth <= 1 ? null : new Releasing(depth - 1, terminalId, role);
+        }
+    }
+
+    /**
+     * The role of the live spawned member occupying {@code terminal} — whether it is currently in
+     * the registry, or mid-teardown between {@link #release} removing its registry entry and
+     * {@link #releaseRemoved} actually stopping its pane (fleetd #702). {@code null} for a terminal
+     * that is neither: this method is the one reader a caller resolver consults before any tab
+     * map, so a live or releasing member's identity never falls back to a tab label.
+     *
+     * <p>Checks the registry directly via {@link #findByTerminal} rather than {@link #roster()},
+     * so this hot-path lookup (consulted on every resolve) never pays for a list copy or a stream.
+     */
+    public MemberRole spawnedMemberRole(String terminal) {
+        MemberSession session = findByTerminal(terminal);
+        if (session != null) {
+            return session.role();
+        }
+        if (terminal == null) {
+            return null;
+        }
+        for (Releasing r : releasing.values()) {
+            if (terminal.equals(r.terminalId())) {
+                return r.role();
+            }
+        }
+        return null;
     }
 
     private void releaseRemoved(String paneId, MemberSession removed, PeerHandle removedHandle,
@@ -382,6 +470,12 @@ public final class SessionManager implements TurnListener {
                 MemberSession resolved = resolveAgentSessionId(removed, removedHandle);
                 notifyReleased(new ReleaseDetail(resolved.terminalId(), resolved.worktree(),
                         resolved.branch(), snapshotRef, resolved.agentSessionId()));
+                String terminal = removed.terminalId();
+                if (terminal != null && !terminal.isBlank()) {
+                    // Without this, a terminal stays marked present after its pane is gone, so a
+                    // later send to the same id would read as deliverable instead of refused.
+                    presence.forget(terminal);
+                }
             }
         }
         // CB-581: the pane must always stop, even if the dirty check above threw. A session removed
@@ -725,6 +819,10 @@ public final class SessionManager implements TurnListener {
                 handle.charterReceipt(),
                 handle.agentSessionId());
         registry.put(handle.id(), session);
+        // A presence contact that already arrived for this terminal found no registry entry to
+        // transition and gave up silently. Retry it now that one exists; remove this call and
+        // such a session stays in SPAWNING even though it is present.
+        reconcilePresence(handle.terminalId());
         handles.put(handle.id(), handle);
         log.debug("acquired worktree session id={} terminal={} profile={} branch={} path={}",
                 handle.id(), handle.terminalId(), session.profile(), session.branch(), session.worktree());
@@ -899,6 +997,20 @@ public final class SessionManager implements TurnListener {
         transitionByTerminal(terminalId, MemberSession.State.SPAWNING, MemberSession.State.READY);
     }
 
+    /**
+     * Completes a newly registered session's {@code SPAWNING -> READY} transition when {@code
+     * terminalId} was already marked present before this ran. A terminal never marked present is
+     * left in {@code SPAWNING}; it reaches {@code READY} normally through {@link #onReady} once
+     * its own contact arrives. Callers must run this only once the session's registry entry is
+     * already visible — {@link #onReady}'s transition matches against that entry, and reconciling
+     * before the entry exists finds nothing to transition.
+     */
+    private void reconcilePresence(String terminalId) {
+        if (terminalId != null && !terminalId.isBlank() && presence.isPresent(terminalId)) {
+            onReady(terminalId);
+        }
+    }
+
     /**
      * Lifecycle hook: a message was delivered into the worker — it is now busy on a turn.
      * The turn count is bumped and the activity timestamp is refreshed. A {@code DONE} session

fleetd/src/test/java/dev/ltms/fleet/FleetDeliverabilityTest.java

@@ -14,10 +14,12 @@ import static org.junit.jupiter.api.Assertions.assertTrue;
 
 /**
  * CB-534: the injector's readiness gate must open for a lead as well as for a present worker.
+ * fleetd #669 follow-up: the same gate must also open for a collaborator, which — like a lead —
+ * is never enrolled in {@link MemberPresence} and never discovered by the lead scan.
  *
- * <p>The bug these cover was silent and slow: a lead was never marked present (only workers are), so
- * every lead→lead delivery sat on the gate for the full readiness grace and failed ~60s later without
- * a keystroke ever reaching the pane.
+ * <p>The bug these cover was silent and slow: a lead (and later a collaborator) was never marked
+ * present (only workers are) and never counted as a lead, so every send to one sat on the gate for
+ * the full readiness grace and failed ~60s later without a keystroke ever reaching the pane.
  */
 class FleetDeliverabilityTest {
 
@@ -25,19 +27,25 @@ class FleetDeliverabilityTest {
         return () -> m;
     }
 
+    private static Supplier<Map<String, String>> collaborators(Map<String, String> m) {
+        return () -> m;
+    }
+
     @Test
     @DisplayName("a worker that has connected its MCP is deliverable")
     void presentWorkerIsDeliverable() {
         MemberPresence presence = new MemberPresence();
         presence.markPresent("term_worker");
 
-        assertTrue(Fleetd.deliverableTo(presence, leads(Map.of())).test("term_worker"));
+        assertTrue(Fleetd.deliverableTo(presence, leads(Map.of()), collaborators(Map.of()))
+                .test("term_worker"));
     }
 
     @Test
     @DisplayName("a worker still in its boot window is held back")
     void absentWorkerIsNotDeliverable() {
-        assertFalse(Fleetd.deliverableTo(new MemberPresence(), leads(Map.of())).test("term_booting"));
+        assertFalse(Fleetd.deliverableTo(new MemberPresence(), leads(Map.of()), collaborators(Map.of()))
+                .test("term_booting"));
     }
 
     @Test
@@ -45,19 +53,31 @@ class FleetDeliverabilityTest {
     void leadIsDeliverableWithoutPresence() {
         MemberPresence presence = new MemberPresence();
         Predicate<String> deliverable =
-                Fleetd.deliverableTo(presence, leads(Map.of("term_lead", "opus-5.0")));
+                Fleetd.deliverableTo(presence, leads(Map.of("term_lead", "opus-5.0")), collaborators(Map.of()));
 
         assertFalse(presence.isPresent("term_lead"), "a lead is never enrolled in worker presence");
         assertTrue(deliverable.test("term_lead"), "…and must be deliverable anyway");
     }
 
     @Test
-    @DisplayName("an unknown terminal is deliverable to neither")
+    @DisplayName("a collaborator is deliverable without ever being marked present or scanned as a lead")
+    void collaboratorIsDeliverableWithoutPresenceOrLeadStatus() {
+        MemberPresence presence = new MemberPresence();
+        Predicate<String> deliverable = Fleetd.deliverableTo(presence, leads(Map.of()),
+                collaborators(Map.of("term_collab", "kevin")));
+
+        assertFalse(presence.isPresent("term_collab"), "a collaborator is never enrolled in worker presence");
+        assertTrue(deliverable.test("term_collab"), "…and must be deliverable anyway");
+    }
+
+    @Test
+    @DisplayName("an unknown terminal is deliverable to none of presence, leads, or collaborators")
     void strangerIsNotDeliverable() {
         MemberPresence presence = new MemberPresence();
         presence.markPresent("term_worker");
 
-        assertFalse(Fleetd.deliverableTo(presence, leads(Map.of("term_lead", "opus-5.0")))
+        assertFalse(Fleetd.deliverableTo(presence, leads(Map.of("term_lead", "opus-5.0")),
+                        collaborators(Map.of("term_collab", "kevin")))
                 .test("term_stranger"));
     }
 
@@ -65,22 +85,47 @@ class FleetDeliverabilityTest {
     @DisplayName("a lead discovered after startup becomes deliverable with no restart")
     void leadSetIsReadThroughOnEveryCall() {
         Map<String, String> discovered = new HashMap<>();
-        Predicate<String> deliverable = Fleetd.deliverableTo(new MemberPresence(), leads(discovered));
+        Predicate<String> deliverable =
+                Fleetd.deliverableTo(new MemberPresence(), leads(discovered), collaborators(Map.of()));
 
         assertFalse(deliverable.test("term_late"));
         discovered.put("term_late", "gpt-sol-5.6"); // leadScan picks up a newly labelled tab
         assertTrue(deliverable.test("term_late"), "the supplier must be re-read, not snapshotted");
     }
 
+    @Test
+    @DisplayName("a collaborator discovered after startup becomes deliverable with no restart")
+    void collaboratorSetIsReadThroughOnEveryCall() {
+        Map<String, String> discovered = new HashMap<>();
+        Predicate<String> deliverable =
+                Fleetd.deliverableTo(new MemberPresence(), leads(Map.of()), collaborators(discovered));
+
+        assertFalse(deliverable.test("term_late_collab"));
+        discovered.put("term_late_collab", "kevin"); // the same tab scan picks up a newly labelled collaborator tab
+        assertTrue(deliverable.test("term_late_collab"), "the supplier must be re-read, not snapshotted");
+    }
+
     @Test
     @DisplayName("forgetting a torn-down worker does not strip a lead of its deliverability")
     void forgetDoesNotDisarmALead() {
         MemberPresence presence = new MemberPresence();
         Predicate<String> deliverable =
-                Fleetd.deliverableTo(presence, leads(Map.of("term_lead", "opus-5.0")));
+                Fleetd.deliverableTo(presence, leads(Map.of("term_lead", "opus-5.0")), collaborators(Map.of()));
 
         presence.forget("term_lead"); // the injector's cleanup path runs against every target
 
         assertTrue(deliverable.test("term_lead"));
     }
+
+    @Test
+    @DisplayName("forgetting a torn-down worker does not strip a collaborator of its deliverability")
+    void forgetDoesNotDisarmACollaborator() {
+        MemberPresence presence = new MemberPresence();
+        Predicate<String> deliverable = Fleetd.deliverableTo(presence, leads(Map.of()),
+                collaborators(Map.of("term_collab", "kevin")));
+
+        presence.forget("term_collab"); // the injector's cleanup path runs against every target
+
+        assertTrue(deliverable.test("term_collab"));
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/FleetdAssemblyCollaboratorHerdrRoutingTest.java

@@ -0,0 +1,165 @@
+package dev.ltms.fleet;
+
+import dev.ltms.fleet.config.ConfigRef;
+import dev.ltms.fleet.config.FleetConfig;
+import dev.ltms.fleet.guard.SubscriptionGuard;
+import dev.ltms.fleet.herdr.FakeHerdr;
+import dev.ltms.fleet.herdr.HerdrClient;
+import dev.ltms.fleet.msg.ReplyInbox;
+import io.javalin.Javalin;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.io.TempDir;
+
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.LinkedHashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.concurrent.Executors;
+import java.util.concurrent.ScheduledExecutorService;
+import java.util.function.LongSupplier;
+
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+import static org.junit.jupiter.api.Assertions.assertSame;
+
+/**
+ * fleetd #669 Unit E. Reaches the real {@link dev.ltms.fleet.herdr.HerdrRouter} that {@link
+ * FleetdAssembly#assembleAndStart} builds and wires — not a copy built for this test — and proves
+ * that a configured collaborator's terminal routes to the LEAD herdr daemon.
+ *
+ * <p>Two distinct {@link FakeHerdr} instances are required, the same pattern {@code
+ * FleetdAssemblyConnectionIdentityTest} and {@code FleetdLeadRolloverAssemblyTest} already use:
+ * with one client shared between {@code herdrSocket} and {@code memberHerdrSocket},
+ * {@code HerdrRouter} folds {@code leadAgents} and {@code memberAgents} into the same instance
+ * (see its constructor), and {@code agentsFor} would return that one object regardless of whether
+ * the collaborator map was ever consulted — invisible to a mutation of the predicate this ticket
+ * fixes. This test's two sockets resolve to two different fakes, so the assertion only passes when
+ * the collaborator's terminal is actually recognised and routed to the lead one.
+ */
+class FleetdAssemblyCollaboratorHerdrRoutingTest {
+
+    private static final Path LEAD_SOCKET = Path.of("/fake/lead-herdr.sock");
+    private static final Path MEMBER_SOCKET = Path.of("/fake/member-herdr.sock");
+
+    private static final class RecordingResourcePorts implements ResourcePorts {
+        final Map<Path, HerdrClient> herdrsBySocket = new LinkedHashMap<>();
+        Runnable shutdownHook;
+
+        @Override
+        public Map<String, String> environment() {
+            return Map.of();
+        }
+
+        @Override
+        public HerdrClient connectHerdr(Path socketPath) {
+            HerdrClient client = herdrsBySocket.get(socketPath);
+            if (client == null) {
+                throw new IllegalStateException("no fake herdr registered for socket " + socketPath);
+            }
+            return client;
+        }
+
+        @Override
+        public Fleetd.AmqpOpener replyInboxOpener() {
+            return (uri, prefetch) -> new ReplyInbox() {
+                @Override public void own(String target) { }
+                @Override public void release(String target) { }
+                @Override public void publish(String target, String msgId, String content) { }
+                @Override public List<InboxMessage> peek(String target) { return List.of(); }
+                @Override public boolean ack(String target, String msgId) { return false; }
+            };
+        }
+
+        @Override
+        public Fleetd.LeadMailboxOpener leadMailboxOpener() {
+            return (uri, selfCoordId, prefetch) -> {
+                throw new UnsupportedOperationException(
+                        "leadMailboxOpener must not be called — no coordinator: block is configured");
+            };
+        }
+
+        @Override
+        public LongSupplier nanoClock() {
+            return System::nanoTime;
+        }
+
+        @Override
+        public LongSupplier wallClockNanos() {
+            return System::nanoTime;
+        }
+
+        @Override
+        public ScheduledExecutorService newScheduler(String purpose) {
+            return Executors.newSingleThreadScheduledExecutor();
+        }
+
+        @Override
+        public void addShutdownHook(Runnable hook) {
+            shutdownHook = hook;
+        }
+
+        @Override
+        public void startHttp(Javalin app, String host, int port) {
+            // Do not bind a real port in this assembly test.
+        }
+
+        @Override
+        public Runnable herdrPollWait() {
+            return () -> {
+                throw new UnsupportedOperationException("FakeHerdr is healthy; no poll wait is expected");
+            };
+        }
+    }
+
+    private static FleetConfig writeConfig(Path dir) throws Exception {
+        Path file = dir.resolve("fleetd.yaml");
+        Files.writeString(file, """
+                bind:
+                  host: 127.0.0.1
+                  port: 8765
+                herdrSocket: "%s"
+                memberHerdrSocket: "%s"
+                idleSleepGuard:
+                  enabled: false
+                fleet:
+                  collaborators:
+                    reviewer-alex:
+                      tab: "collab: alex"
+                profiles:
+                  sonnet:
+                    subscription: true
+                    argv: ["ccs", "sonnet"]
+                """.formatted(LEAD_SOCKET, MEMBER_SOCKET));
+        return FleetConfig.load(file);
+    }
+
+    @Test
+    void assembledRouterRoutesACollaboratorTerminalToTheLeadDaemon(@TempDir Path dir) throws Exception {
+        FleetConfig cfg = writeConfig(dir);
+        RecordingResourcePorts ports = new RecordingResourcePorts();
+
+        // The fixed FakeHerdr fixture already ties terminal "term_a" to a live agent on tab
+        // "w2:t7" (pane "w2:p7") — seeding only the tab LABEL to match the configured collaborator
+        // is enough to make LeadTabScanner resolve "term_a" as that collaborator. Seeded on the
+        // LEAD fake only: a collaborator's pane lives in the lead daemon, exactly like a lead's.
+        FakeHerdr lead = new FakeHerdr().withTab("w2", "w2:t7", "collab: alex");
+        FakeHerdr member = new FakeHerdr();
+        ports.herdrsBySocket.put(LEAD_SOCKET, lead);
+        ports.herdrsBySocket.put(MEMBER_SOCKET, member);
+
+        FleetdRuntime runtime = FleetdAssembly.assembleAndStart(new AssemblyInputs(cfg,
+                new ConfigRef(dir.resolve("fleetd.yaml"), cfg), new SubscriptionGuard(cfg.guard().hostSet())), ports);
+        try {
+            assertSame(runtime.router().leadAgents(), runtime.router().agentsFor("term_a"),
+                    "a configured collaborator's terminal must route to the LEAD daemon — "
+                            + "FleetdAssembly must wire the collaborator map into the router's "
+                            + "predicate, not just LeadTabScanner.get()");
+            assertSame(runtime.router().memberAgents(), runtime.router().agentsFor("term_shell"),
+                    "control: a terminal naming neither a lead nor a collaborator (term_shell, on "
+                            + "the unlabelled tab w2:t8) must still route to the member daemon");
+        } finally {
+            assertNotNull(ports.shutdownHook, "control: assembly must capture its shutdown hook");
+            ports.shutdownHook.run();
+        }
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/FleetdAssemblyLeadTabScannerExclusionTest.java

@@ -31,8 +31,7 @@ import static org.junit.jupiter.api.Assertions.assertTrue;
 
 /**
  * fleetd #670 — pins the {@code excludedWorkspaceLabels} argument {@link FleetdAssembly}'s
- * production boot path passes to {@link LeadTabScanner} at {@code FleetdAssembly.java:265}
- * ({@code Set.of()}).
+ * production boot path passes to {@link LeadTabScanner} ({@code Set.of()}).
  *
  * <p>{@code LeadTabScannerTest} already covers this constructor parameter, but it builds its own
  * {@link LeadTabScanner} with its own set, so it tests the seam and proves nothing about the
@@ -191,7 +190,7 @@ class FleetdAssemblyLeadTabScannerExclusionTest {
             Set<?> excluded = (Set<?>) excludedField.get(leads);
 
             assertTrue(excluded.isEmpty(),
-                    "FleetdAssembly.java:265 must pass an empty excludedWorkspaceLabels to "
+                    "FleetdAssembly must pass an empty excludedWorkspaceLabels to "
                             + "LeadTabScanner — scanning member tabs would demote the lead to a worker");
         } finally {
             assertNotNull(ports.shutdownHook, "control: assembly must capture its shutdown hook");

fleetd/src/test/java/dev/ltms/fleet/FleetdAssemblyTimingDefaultsTest.java

@@ -0,0 +1,190 @@
+package dev.ltms.fleet;
+
+import dev.ltms.fleet.config.ConfigRef;
+import dev.ltms.fleet.config.FleetConfig;
+import dev.ltms.fleet.guard.SubscriptionGuard;
+import dev.ltms.fleet.herdr.FakeHerdr;
+import dev.ltms.fleet.herdr.HerdrClient;
+import dev.ltms.fleet.inject.Injector;
+import dev.ltms.fleet.inject.StatusPoller;
+import dev.ltms.fleet.msg.LeadChannelHandle;
+import dev.ltms.fleet.msg.LeadCoordLoop;
+import dev.ltms.fleet.msg.LeadMessage;
+import dev.ltms.fleet.msg.ReplyInbox;
+import dev.ltms.fleet.msg.ReplyPushLoop;
+import io.javalin.Javalin;
+import org.junit.jupiter.api.AfterEach;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.io.TempDir;
+
+import java.lang.reflect.Field;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.List;
+import java.util.Map;
+import java.util.concurrent.Executors;
+import java.util.concurrent.ScheduledExecutorService;
+import java.util.function.LongSupplier;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+
+/**
+ * Asserts that the assembled loops use the production reminder, coordination, and delivery timing
+ * defaults when no {@code primary:} block configures the reply-push values.
+ */
+class FleetdAssemblyTimingDefaultsTest {
+
+    private static final class FakeLeadChannel implements LeadChannelHandle {
+        @Override
+        public void publish(String toCoordId, LeadMessage message) {
+        }
+
+        @Override
+        public List<LeadMessage> peek() {
+            return List.of();
+        }
+
+        @Override
+        public void ack(String msgId) {
+        }
+
+        @Override
+        public String selfCoordId() {
+            return "test-lead";
+        }
+
+        @Override
+        public boolean heldDurable() {
+            return true;
+        }
+
+        @Override
+        public MailboxState inspect(String coordId) {
+            return MailboxState.unknown(coordId);
+        }
+
+        @Override
+        public void close() {
+        }
+    }
+
+    private static final class TestResourcePorts implements ResourcePorts {
+        final FakeHerdr herdr = new FakeHerdr();
+        Runnable shutdownHook;
+
+        @Override
+        public Map<String, String> environment() {
+            return Map.of();
+        }
+
+        @Override
+        public HerdrClient connectHerdr(Path socketPath) {
+            return herdr;
+        }
+
+        @Override
+        public Fleetd.AmqpOpener replyInboxOpener() {
+            return (uri, prefetch) -> new ReplyInbox() {
+                @Override public void own(String target) { }
+                @Override public void release(String target) { }
+                @Override public void publish(String target, String msgId, String content) { }
+                @Override public List<InboxMessage> peek(String target) { return List.of(); }
+                @Override public boolean ack(String target, String msgId) { return false; }
+            };
+        }
+
+        @Override
+        public Fleetd.LeadMailboxOpener leadMailboxOpener() {
+            return (uri, selfCoordId, prefetch) -> new FakeLeadChannel();
+        }
+
+        @Override
+        public LongSupplier nanoClock() {
+            return System::nanoTime;
+        }
+
+        @Override
+        public LongSupplier wallClockNanos() {
+            return System::nanoTime;
+        }
+
+        @Override
+        public ScheduledExecutorService newScheduler(String purpose) {
+            return Executors.newSingleThreadScheduledExecutor();
+        }
+
+        @Override
+        public void addShutdownHook(Runnable hook) {
+            shutdownHook = hook;
+        }
+
+        @Override
+        public void startHttp(Javalin app, String host, int port) {
+        }
+
+        @Override
+        public Runnable herdrPollWait() {
+            return () -> {
+                throw new UnsupportedOperationException("FakeHerdr is healthy; no poll wait is expected");
+            };
+        }
+    }
+
+    private TestResourcePorts ports;
+
+    @AfterEach
+    void tearDown() {
+        if (ports != null && ports.shutdownHook != null) {
+            ports.shutdownHook.run();
+        }
+    }
+
+    private static FleetConfig writeConfig(Path dir) throws Exception {
+        Path file = dir.resolve("fleetd.yaml");
+        Files.writeString(file, """
+                bind:
+                  host: 127.0.0.1
+                  port: 8765
+                idleSleepGuard:
+                  enabled: false
+                coordinator:
+                  uri: "amqp://fake-lead-broker/vh"
+                  selfId: "test-lead"
+                """);
+        return FleetConfig.load(file);
+    }
+
+    private FleetdRuntime assemble(Path dir) throws Exception {
+        FleetConfig cfg = writeConfig(dir);
+        ports = new TestResourcePorts();
+        return FleetdAssembly.assembleAndStart(new AssemblyInputs(cfg,
+                new ConfigRef(dir.resolve("fleetd.yaml"), cfg), new SubscriptionGuard(cfg.guard().hostSet())), ports);
+    }
+
+    private static long longField(Object target, String name) throws Exception {
+        Field field = target.getClass().getDeclaredField(name);
+        field.setAccessible(true);
+        return field.getLong(target);
+    }
+
+    @Test
+    void productionBootPathUsesTheExpectedLoopTimingDefaults(@TempDir Path dir) throws Exception {
+        FleetdRuntime runtime = assemble(dir);
+
+        ReplyPushLoop pushLoop = runtime.pushLoop();
+        assertEquals(5, longField(pushLoop, "maxReminders"),
+                "without primary:, ReplyPushLoop must stop after five reminder attempts");
+        assertEquals(15_000L, longField(pushLoop, "backoffMs"),
+                "without primary:, ReplyPushLoop must wait fifteen seconds before the next reminder");
+
+        LeadCoordLoop leadCoordLoop = runtime.leadCoordLoop();
+        assertNotNull(leadCoordLoop, "control: coordinator: must build LeadCoordLoop");
+        assertEquals(3_000L, longField(leadCoordLoop, "intervalMs"),
+                "LeadCoordLoop must poll for peer-lead mail every three seconds");
+
+        StatusPoller poller = runtime.poller();
+        assertEquals(Injector.POLL_INTERVAL_MILLIS, longField(poller, "intervalMillis"),
+                "StatusPoller must use Injector's delivery poll interval");
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/FleetdAssemblyTurnRegistrarBehaviouralTest.java

@@ -125,6 +125,7 @@ class FleetdAssemblyTurnRegistrarBehaviouralTest {
                     primary:
                       tab: "lead: primary"
                       profile: sonnet
+                      workspace: "ltms"
                 profiles:
                   sonnet:
                     subscription: true

fleetd/src/test/java/dev/ltms/fleet/FleetdBackendErrorSinkTest.java

@@ -91,6 +91,11 @@ class FleetdBackendErrorSinkTest {
                 return MAPPER.createObjectNode().set("agent", MAPPER.createObjectNode()
                         .put("terminal_id", "term_primary").put("agent_status", "idle"));
             }
+            if ("agent.read".equals(method)) {
+                // The lead-nudge paths read the input box before pasting into it.
+                return MAPPER.createObjectNode().set("read",
+                        MAPPER.createObjectNode().put("text", FakeHerdr.IDLE_PROMPT_CARET));
+            }
             if ("agent.prompt".equals(method)) {
                 prompts.add(params);
                 sendLatch.countDown();

fleetd/src/test/java/dev/ltms/fleet/FleetdLeadContextSourceWindowAssemblyTest.java

@@ -171,6 +171,7 @@ class FleetdLeadContextSourceWindowAssemblyTest {
                     %s:
                       tab: "%s"
                       profile: %s
+                      workspace: "ltms"
                 profiles:
                   %s:
                     subscription: true

fleetd/src/test/java/dev/ltms/fleet/FleetdLeadRolloverAssemblyTest.java

@@ -6,6 +6,8 @@ import dev.ltms.fleet.guard.SubscriptionGuard;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.FakeHerdr;
 import dev.ltms.fleet.herdr.HerdrClient;
+import dev.ltms.fleet.herdr.WorkspaceControl;
+import dev.ltms.fleet.lead.LeadLauncher;
 import dev.ltms.fleet.lead.LeadRollover;
 import dev.ltms.fleet.msg.ReplyInbox;
 import io.javalin.Javalin;
@@ -34,7 +36,7 @@ import static org.junit.jupiter.api.Assertions.fail;
  * fleetd #612 Unit A) with three methods: {@code unrelatedAnchorStillPresent} (a scaffold anchor,
  * not an independent claim — needs no replacement of its own), {@code
  * mainStillCallsTheLeadRolloverFactory} (the call-site pin replaced by {@link
- * #assembledLeadRolloverRunsTheRealClearAndBootstrapSequence}), and {@code
+ * #assembledLeadRolloverEndsTheOldPaneThroughTheRealHerdrRouter}), and {@code
  * factoryGatesOnConfigPresence} (the absent-config claim replaced by {@link
  * #absentLeadRolloverConfigMeansNoRolloverIsBuilt} — a claim this ticket found was NOT actually
  * covered behaviourally anywhere else: {@code LeadRolloverTest}'s only related assertion is
@@ -50,8 +52,8 @@ import static org.junit.jupiter.api.Assertions.fail;
  * invisible to this test, even though the two are genuinely different daemons in production. This
  * version configures two distinct sockets and two distinct {@link FakeHerdr} instances (the same
  * pattern {@code FleetdAssemblyConnectionIdentityTest}, fleetd #612 B2, already uses to separate
- * lead from member) and asserts the roll's {@code /clear}/bootstrap sends land on the LEAD fake
- * and never on the MEMBER one.
+ * lead from member) and asserts the roll's {@code pane.close} call lands on the LEAD fake and
+ * never on the MEMBER one.
  */
 class FleetdLeadRolloverAssemblyTest {
 
@@ -170,6 +172,7 @@ class FleetdLeadRolloverAssemblyTest {
                     opus:
                       tab: "lead: opus"
                       cwd: "%s"
+                      workspace: "ltms"
                 leadRollover:
                   handoverPath: handover.md
                   requireOperatorConfirm: false
@@ -177,11 +180,48 @@ class FleetdLeadRolloverAssemblyTest {
         return FleetConfig.load(f);
     }
 
-    @SuppressWarnings("unchecked")
+    /**
+     * Unlike {@link #writeConfig}, this names a {@code profile:} for the lead and declares it
+     * under {@code profiles:}, so {@code LeadLauncher#relaunch} can actually start a fresh agent
+     * instead of refusing with "names no profile". {@code relaunchReadySeconds} is cut to 2s so
+     * the recognition wait (expected to time out — see the test) does not cost real test seconds.
+     */
+    private static FleetConfig writeConfigWithRelaunchableLead(Path dir, Path leadCwd) throws Exception {
+        Path f = dir.resolve("fleetd.yaml");
+        Files.writeString(f, """
+                bind:
+                  host: 127.0.0.1
+                  port: 8765
+                herdrSocket: "%s"
+                memberHerdrSocket: "%s"
+                idleSleepGuard:
+                  enabled: false
+                broker:
+                  uri: "amqp://fake-test-broker/vh"
+                fleet:
+                  leaders:
+                    opus:
+                      tab: "lead: opus"
+                      cwd: "%s"
+                      profile: opus
+                      workspace: "ltms"
+                profiles:
+                  opus:
+                    subscription: true
+                    argv: ["ccs", "opus"]
+                leadRollover:
+                  handoverPath: handover.md
+                  requireOperatorConfirm: false
+                  relaunchReadySeconds: 2
+                """.formatted(LEAD_SOCKET, MEMBER_SOCKET, leadCwd.toString()));
+        return FleetConfig.load(f);
+    }
+
     @Test
-    @DisplayName("[BEHAVIOURAL] the real assembled LeadRollover runs the full open/confirm/continuation "
-            + "sequence — /clear, then bootstrapText — through the real herdr router")
-    void assembledLeadRolloverRunsTheRealClearAndBootstrapSequence(@TempDir Path dir) throws Exception {
+    @DisplayName("[BEHAVIOURAL] the real assembled LeadRollover runs the open/confirm/continuation "
+            + "sequence through the real herdr router — ending the old pane, then giving up once it "
+            + "never reports gone")
+    void assembledLeadRolloverEndsTheOldPaneThroughTheRealHerdrRouter(@TempDir Path dir) throws Exception {
         Path leadCwd = dir.resolve("lead-workspace");
         Files.createDirectories(leadCwd);
         FleetConfig cfg = writeConfig(dir, leadCwd);
@@ -204,6 +244,11 @@ class FleetdLeadRolloverAssemblyTest {
                 + "Fleetd.leadRollover(...) call site — a mutation to `LeadRollover leadRollover = "
                 + "null;` at that call site can never pass this");
 
+        // assembleAndStart's own boot work (the orphan-worker reap) makes a real call on the
+        // member daemon before the roll ever starts. Clear it here so the assertion below measures
+        // only what the roll itself does, not what daemon startup does.
+        member.calls.clear();
+
         LeadRollover.PendingRollover pending = rollover.open("term_a", "fleetd #612 B3 test");
         String expectedHandoverPath = leadCwd.resolve("handover.md").normalize().toString();
         assertEquals(expectedHandoverPath, pending.handoverPath());
@@ -223,40 +268,109 @@ class FleetdLeadRolloverAssemblyTest {
         // constructor), so this polls the real FleetMcp.leadRollover() instance's status(token)
         // until the real continuation finishes.
         LeadRollover.RollStatus status = pollUntilTerminal(rollover, pending.token());
-        assertEquals(LeadRollover.RollState.ROLLED, status.state(),
-                "the full happy path must complete: FakeHerdr's default agent status is 'idle', so "
-                        + "the turn-boundary wait settles immediately and the post-/clear wait "
-                        + "releases via its pickup-grace path — detail: " + status.detail());
 
-        // Prove the real herdr router actually sent BOTH messages, in order, to the real LEAD
-        // pane — this is the one thing a source-text pin on the call site could never show.
+        // FakeHerdr's pane.get is a fixed canned response that never reports a pane as gone, so the
+        // real router's death poll runs out its whole budget and the roll stops here — proving the
+        // real teardown call landed on the real LEAD pane without ever reaching a relaunch or a send.
+        assertEquals(LeadRollover.RollState.OLD_PANE_NEVER_DIED, status.state(),
+                "the old pane never reports gone against this fake, so the roll must stop with "
+                        + "OLD_PANE_NEVER_DIED rather than ever relaunching or sending anything — "
+                        + "detail: " + status.detail());
+
+        // Prove the real herdr router actually closed the real LEAD pane — this is the one thing a
+        // source-text pin on the call site could never show.
+        boolean closedOldPane = lead.calls.stream()
+                .anyMatch(c -> c.method().equals("pane.close")
+                        && c.params() instanceof Map<?, ?> m && "w2:p7".equals(m.get("pane_id")));
+        assertTrue(closedOldPane, "endOldSession must close the real old pane (w2:p7) through the "
+                + "real LEAD herdr client, got calls: " + lead.calls);
+
+        // No agent.prompt is ever sent on this path: the roll stops at the pane-death wait, strictly
+        // before the relaunch and the final send step.
         List<FakeHerdr.Call> prompts = lead.calls.stream()
                 .filter(c -> c.method().equals("agent.prompt"))
                 .toList();
-        assertTrue(prompts.size() >= 2, "expected at least a /clear send and a bootstrapText send "
-                + "on the LEAD daemon, got " + prompts.size() + " agent.prompt calls: " + prompts);
-        assertEquals("/clear", ((Map<String, Object>) prompts.get(0).params()).get("text"),
-                "the first send must be the literal /clear housekeeping command");
-        Object secondText = ((Map<String, Object>) prompts.get(1).params()).get("text");
-        assertTrue(secondText instanceof String && ((String) secondText).contains(expectedHandoverPath),
-                "the second send must be the default bootstrapText naming the resolved handover "
-                        + "path, got: " + secondText);
+        assertTrue(prompts.isEmpty(), "a roll that stops at OLD_PANE_NEVER_DIED must never reach the "
+                + "send step, got agent.prompt call(s) on the LEAD daemon: " + prompts);
 
         // fleetd #612 B3 correction: prove the roll never touches the MEMBER daemon. A mutation
         // swapping router.leadAgents() for router.memberAgents() at the real call site would move
-        // both sends above onto `member` instead, which this assertion catches — the thing the
-        // single-fake version of this test could never see, because both wrapped the same client.
+        // the pane.close call above onto `member` instead, which this assertion catches — the thing
+        // the single-fake version of this test could never see, because both wrapped the same client.
+        assertTrue(member.calls.isEmpty(), "the roll must be wired to the LEAD daemon only — got "
+                + member.calls.size() + " call(s) recorded on the MEMBER daemon since the roll began: "
+                + member.calls);
+    }
+
+    /**
+     * Exercises the relaunch site {@link #assembledLeadRolloverEndsTheOldPaneThroughTheRealHerdrRouter}
+     * never reaches: with the old pane confirmed gone, the roll relaunches a fresh lead, and
+     * {@code bootstrapText} must reach it even though recognition times out (FakeHerdr's
+     * {@code tab.list} is a fixed canned response that never reflects the relaunch's own
+     * {@code tab.rename}, so the fresh terminal is never recognised as a live lead). Same
+     * dual-socket shape as the sibling test: two distinct {@link FakeHerdr} instances, so a
+     * {@code bootstrapText} send wired to the wrong daemon is visible.
+     */
+    @Test
+    @DisplayName("[BEHAVIOURAL] bootstrapText reaches the fresh LEAD terminal even when recognition "
+            + "times out, and the MEMBER daemon never sees it")
+    void bootstrapTextReachesTheFreshLeadTerminalEvenWhenRecognitionTimesOut(@TempDir Path dir) throws Exception {
+        Path leadCwd = dir.resolve("lead-workspace");
+        Files.createDirectories(leadCwd);
+        FleetConfig cfg = writeConfigWithRelaunchableLead(dir, leadCwd);
+        ConfigRef config = new ConfigRef(dir.resolve("fleetd.yaml"), cfg);
+        SubscriptionGuard guard = new SubscriptionGuard(cfg.guard().hostSet());
+        RecordingResourcePorts ports = new RecordingResourcePorts();
+        FakeHerdr lead = new FakeHerdr();
+        lead.withTab("w2", "w2:t7", "lead: opus");
+        // Lets the old pane (w2:p7) report gone once pane.close actually reaches it, so the roll
+        // proceeds to relaunch instead of stopping at OLD_PANE_NEVER_DIED.
+        lead.paneGoneAfterClose("w2:p7");
+        FakeHerdr member = new FakeHerdr();
+        ports.herdrsBySocket.put(LEAD_SOCKET, lead);
+        ports.herdrsBySocket.put(MEMBER_SOCKET, member);
+
+        FleetdRuntime runtime = FleetdAssembly.assembleAndStart(new AssemblyInputs(cfg, config, guard), ports);
+
+        LeadRollover rollover = runtime.mcp().leadRollover();
+        assertNotNull(rollover, "leadRollover: is present in this test's config, so a real "
+                + "LeadRollover must have been built");
+
+        LeadRollover.PendingRollover pending = rollover.open("term_a", "bootstrapText relaunch test");
+        Thread.sleep(50);
+        Files.writeString(Path.of(pending.handoverPath()), "handover content for bootstrapText test");
+
+        LeadRollover.RollDecision decision = rollover.confirm("term_a", pending.token(), true);
+        assertTrue(decision.accepted(), "confirm() must approve — got: " + decision);
+
+        LeadRollover.RollStatus status = pollUntilTerminal(rollover, pending.token());
+        assertEquals(LeadRollover.RollState.RELAUNCH_NOT_RECOGNISED, status.state(),
+                "the fresh terminal is never recognised against this fake's static tab.list, so the "
+                        + "roll must reach RELAUNCH_NOT_RECOGNISED — not an earlier failure state and "
+                        + "not ROLLED — detail: " + status.detail());
+
+        @SuppressWarnings("unchecked")
+        List<FakeHerdr.Call> leadPrompts = lead.calls.stream()
+                .filter(c -> c.method().equals("agent.prompt"))
+                .toList();
+        assertEquals(1, leadPrompts.size(), "exactly one bootstrapText send is expected, on the LEAD "
+                + "daemon, once recognition gives up — got: " + leadPrompts);
+        Object text = ((Map<String, Object>) leadPrompts.get(0).params()).get("text");
+        assertTrue(text instanceof String && ((String) text).contains("handover.md"),
+                "the send must be bootstrapText naming the resolved handover path, got: " + text);
+
+        // Scoped to agent.prompt specifically, not every MEMBER call: the orphan-worker reap also
+        // talks to the MEMBER daemon once, unconditionally, at daemon boot — unrelated to this roll.
         List<FakeHerdr.Call> memberPrompts = member.calls.stream()
                 .filter(c -> c.method().equals("agent.prompt"))
                 .toList();
-        assertTrue(memberPrompts.isEmpty(), "the roll must be wired to the LEAD daemon only — got "
-                + memberPrompts.size() + " agent.prompt call(s) on the MEMBER daemon instead: "
-                + memberPrompts);
+        assertTrue(memberPrompts.isEmpty(), "bootstrapText must never be sent to the MEMBER daemon, "
+                + "got: " + memberPrompts);
     }
 
     private static LeadRollover.RollStatus pollUntilTerminal(LeadRollover rollover, String token)
             throws InterruptedException {
-        long deadline = System.nanoTime() + java.util.concurrent.TimeUnit.SECONDS.toNanos(10);
+        long deadline = System.nanoTime() + java.util.concurrent.TimeUnit.SECONDS.toNanos(15);
         while (System.nanoTime() < deadline) {
             LeadRollover.RollStatus status = rollover.status(token);
             if (status.state() != LeadRollover.RollState.PENDING
@@ -283,8 +397,10 @@ class FleetdLeadRolloverAssemblyTest {
                 """);
         ConfigRef config = new ConfigRef(yaml, FleetConfig.load(yaml));
         AgentControl agents = new AgentControl(new FakeHerdr());
+        WorkspaceControl spaces = new WorkspaceControl(new FakeHerdr());
+        LeadLauncher launcher = new LeadLauncher(agents, spaces, config.get());
 
-        LeadRollover rollover = Fleetd.leadRollover(config.get(), agents, config, Map::of);
+        LeadRollover rollover = Fleetd.leadRollover(config.get(), agents, spaces, launcher, config, Map::of);
 
         assertNull(rollover, "leadRollover: is absent from this config, so the factory's opt-in "
                 + "gate (`if (cfg.leadRollover() == null) return null;`) must fire and no "

fleetd/src/test/java/dev/ltms/fleet/FleetdLeadRolloverWorkspaceLookupTest.java

@@ -4,6 +4,8 @@ import dev.ltms.fleet.config.ConfigRef;
 import dev.ltms.fleet.config.FleetConfig;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.FakeHerdr;
+import dev.ltms.fleet.herdr.WorkspaceControl;
+import dev.ltms.fleet.lead.LeadLauncher;
 import dev.ltms.fleet.lead.LeadRollover;
 import org.junit.jupiter.api.DisplayName;
 import org.junit.jupiter.api.Test;
@@ -54,6 +56,11 @@ class FleetdLeadRolloverWorkspaceLookupTest {
         return new AgentControl(new FakeHerdr());
     }
 
+    /** None of this class's tests reach the deferred continuation, so a plain fake is enough. */
+    private static LeadLauncher fakeLauncher(FleetConfig cfg) {
+        return new LeadLauncher(fakeAgents(), new WorkspaceControl(new FakeHerdr()), cfg);
+    }
+
     @Test
     @DisplayName("[BEHAVIOURAL] Fleetd.leadRollover(...) resolves a relative handoverPath against "
             + "the CALLING lead's configured cwd, not the daemon's own working directory")
@@ -74,7 +81,8 @@ class FleetdLeadRolloverWorkspaceLookupTest {
                 """.formatted(leadCwd.toString()));
         ConfigRef config = new ConfigRef(yaml, FleetConfig.load(yaml));
 
-        LeadRollover rollover = Fleetd.leadRollover(config.get(), fakeAgents(), config,
+        LeadRollover rollover = Fleetd.leadRollover(config.get(), fakeAgents(),
+                new WorkspaceControl(new FakeHerdr()), fakeLauncher(config.get()), config,
                 () -> Map.of("term_opus", "opus"));
         assertNotNull(rollover, "leadRollover: is present in the loaded config, so the factory "
                 + "must construct an object");
@@ -105,7 +113,8 @@ class FleetdLeadRolloverWorkspaceLookupTest {
 
         // No lead has been discovered yet — exactly the real shape of a lead the live tab scan
         // has not yet scanned, or one with no fleet.leaders entry at all.
-        LeadRollover rollover = Fleetd.leadRollover(config.get(), fakeAgents(), config, Map::of);
+        LeadRollover rollover = Fleetd.leadRollover(config.get(), fakeAgents(),
+                new WorkspaceControl(new FakeHerdr()), fakeLauncher(config.get()), config, Map::of);
         assertNotNull(rollover);
 
         LeadRollover.PendingRollover pending = rollover.open("term_unknown", "test");
@@ -144,7 +153,8 @@ class FleetdLeadRolloverWorkspaceLookupTest {
         // below — exactly the natural mistake to make, since leads are discovered by a live tab
         // scan that runs AFTER this factory is constructed at startup.
         Map<String, String> liveLeadTerminals = new HashMap<>();
-        LeadRollover rollover = Fleetd.leadRollover(config.get(), fakeAgents(), config,
+        LeadRollover rollover = Fleetd.leadRollover(config.get(), fakeAgents(),
+                new WorkspaceControl(new FakeHerdr()), fakeLauncher(config.get()), config,
                 () -> liveLeadTerminals);
         assertNotNull(rollover);
 

fleetd/src/test/java/dev/ltms/fleet/FleetdLeadSeatAssemblyTest.java

@@ -143,6 +143,7 @@ class FleetdLeadSeatAssemblyTest {
                     opus:
                       tab: "lead: opus"
                       profile: sonnet
+                      workspace: "ltms"
                 profiles:
                   sonnet:
                     subscription: true

fleetd/src/test/java/dev/ltms/fleet/PackageCyclesTest.java

@@ -1,96 +1,174 @@
 package dev.ltms.fleet;
 
-import com.tngtech.archunit.base.DescribedPredicate;
+import com.tngtech.archunit.core.domain.Dependency;
 import com.tngtech.archunit.core.domain.JavaClass;
-import com.tngtech.archunit.core.domain.JavaClass.Predicates;
+import com.tngtech.archunit.core.domain.JavaClasses;
 import com.tngtech.archunit.core.importer.ClassFileImporter;
 import com.tngtech.archunit.core.importer.ImportOption;
 import com.tngtech.archunit.library.dependencies.SliceRule;
 import com.tngtech.archunit.library.dependencies.SlicesRuleDefinition;
 import org.junit.jupiter.api.Test;
 
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Set;
+import java.util.TreeSet;
+
+import static org.junit.jupiter.api.Assertions.fail;
+
 /**
- * fleetd #131 (CB-627): enforce package boundaries with an ArchUnit test instead of a
- * Maven module split.
+ * Enforces package boundaries between the top-level {@code dev.ltms.fleet.*} packages.
  *
- * <p>This test fails the build the moment a NEW cycle appears between the top-level
- * {@code dev.ltms.fleet.*} packages. Today's cycles are recorded below as explicit,
- * narrow exceptions: each one ignores dependencies between exactly the two named
- * packages, in both directions, and nothing else. A cycle through any other pair of
- * packages -- or a brand new pair -- still fails this test.
+ * <p>{@link #BASELINE_EDGES} names the exact {@code origin class -> target class}
+ * dependencies allowed to cross a top-level package boundary. Any dependency between two
+ * top-level packages that is not in that set fails this test, including a brand new
+ * dependency between a pair of packages that already has other baselined edges. A baseline
+ * entry whose dependency no longer exists in the code also fails this test, so the baseline
+ * always names exactly today's exceptions and nothing more.
  *
  * <p><b>Main code only.</b> The import excludes test classes
- * ({@link ImportOption.Predefined#DO_NOT_INCLUDE_TESTS}). Test code legitimately wires
- * across many packages for setup and mocking; that is not part of the shipped
- * architecture this rule protects. Verified: importing test classes too pulls in a much
- * larger, noisier cycle set -- {@code herdr}, {@code member}, {@code peer}, {@code
- * config}, {@code guard} and {@code placement} all show up in cycles that disappear the
- * moment test classes are excluded. Scanning off the classpath via {@code
- * importPackages(...)} (not a hardcoded {@code target/classes} path) also keeps this
- * test correct regardless of the working directory the build is invoked from.
- *
- * <p><b>No package moves here</b> -- ticket #131 is explicit that removing a cycle is
- * its own, later PR. See the comment on each exception below for which ticket step
- * removes it.
+ * ({@link ImportOption.Predefined#DO_NOT_INCLUDE_TESTS}). Scanning off the classpath via
+ * {@code importPackages(...)} keeps this test correct regardless of the working directory
+ * the build is invoked from.
  */
 class PackageCyclesTest {
 
+    /**
+     * Exact {@code "origin -> target"} class dependencies allowed to cross a top-level
+     * package boundary. Each entry is one directed edge between two specific classes; a
+     * two-way relationship between a pair of packages is listed as two separate entries,
+     * one per direction.
+     */
+    private static final Set<String> BASELINE_EDGES = Set.of(
+            "dev.ltms.fleet.auth.CallerResolver -> dev.ltms.fleet.mcp.ConnectionIdentity",
+            "dev.ltms.fleet.auth.CallerResolver -> dev.ltms.fleet.mcp.ConnectionIdentity$Caller",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.auth.AuditLog",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.auth.Authz",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.auth.Authz$Action",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.auth.CallerResolver",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.auth.Principal",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.auth.Role",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.LeadChannel",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.LeadChannel$MailboxState",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.LeadMessage",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$AskOutcome",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$AskResult",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$Outcome",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$Outstanding",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$PendingAsk",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$Phase",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$Reply",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$ReplyOutcome",
+            "dev.ltms.fleet.mcp.FleetMcp -> dev.ltms.fleet.msg.MessageService$TaskView",
+            "dev.ltms.fleet.mcp.FleetMcp$1 -> dev.ltms.fleet.msg.MessageService$AskOutcome",
+            "dev.ltms.fleet.mcp.FleetMcp$1 -> dev.ltms.fleet.msg.MessageService$Outcome",
+            "dev.ltms.fleet.mcp.FleetMcp$1 -> dev.ltms.fleet.msg.MessageService$Phase",
+            "dev.ltms.fleet.mcp.FleetMcp$CoordinationSource -> dev.ltms.fleet.msg.LeadChannel",
+            "dev.ltms.fleet.msg.LeadHeartbeatLoop -> dev.ltms.fleet.mcp.PrimaryRegistry",
+            "dev.ltms.fleet.msg.ReplyPushLoop -> dev.ltms.fleet.mcp.PrimaryRegistry",
+            "dev.ltms.fleet.inject.CompletionResolver -> dev.ltms.fleet.msg.Rendezvous",
+            "dev.ltms.fleet.inject.CompletionResolver -> dev.ltms.fleet.msg.Rendezvous$Resolution",
+            "dev.ltms.fleet.inject.CompletionResolver -> dev.ltms.fleet.msg.TurnToken",
+            "dev.ltms.fleet.inject.CompletionResolver$InFlight -> dev.ltms.fleet.msg.Rendezvous$Resolution",
+            "dev.ltms.fleet.inject.Injector -> dev.ltms.fleet.msg.TurnToken",
+            "dev.ltms.fleet.inject.Injector$Pending -> dev.ltms.fleet.msg.TurnToken",
+            "dev.ltms.fleet.inject.TurnListener -> dev.ltms.fleet.msg.TurnToken",
+            "dev.ltms.fleet.inject.TurnRegistrar -> dev.ltms.fleet.msg.TurnToken",
+            "dev.ltms.fleet.msg.MessageService -> dev.ltms.fleet.inject.Injector",
+            "dev.ltms.fleet.msg.MessageService -> dev.ltms.fleet.inject.Injector$Cancellation",
+            "dev.ltms.fleet.msg.MessageService -> dev.ltms.fleet.inject.Injector$Delivery",
+            "dev.ltms.fleet.metrics.FleetMetrics -> dev.ltms.fleet.msg.ReplyInbox",
+            "dev.ltms.fleet.msg.LeadHeartbeatLoop -> dev.ltms.fleet.metrics.Metrics",
+            "dev.ltms.fleet.msg.MessageService -> dev.ltms.fleet.metrics.Metrics",
+            "dev.ltms.fleet.msg.ReplyPushLoop -> dev.ltms.fleet.metrics.Metrics",
+            "dev.ltms.fleet.msg.LeadHeartbeatLoop -> dev.ltms.fleet.session.MemberSession",
+            "dev.ltms.fleet.msg.LeadHeartbeatLoop -> dev.ltms.fleet.session.MemberSession$State",
+            "dev.ltms.fleet.session.SessionManager -> dev.ltms.fleet.msg.TurnToken"
+    );
+
+    private static final String ROOT_PACKAGE = "dev.ltms.fleet.";
+
     @Test
     void packagesAreFreeOfCycles() {
-        var classes = new ClassFileImporter()
+        JavaClasses classes = new ClassFileImporter()
                 .withImportOption(ImportOption.Predefined.DO_NOT_INCLUDE_TESTS)
                 .importPackages("dev.ltms.fleet");
 
+        checkBaselineMatchesTodaysEdges(classes);
+
         SliceRule rule = SlicesRuleDefinition.slices()
                 .matching("dev.ltms.fleet.(*)..")
                 .should().beFreeOfCycles();
-
-        // fleetd #131 step 1: move ConnectionIdentity so authz stops depending on the
-        // MCP layer. Evidence: auth/CallerResolver.java:3 imports mcp.ConnectionIdentity;
-        // mcp/FleetMcp.java:3-7 imports auth.AuditLog, Authz, CallerResolver, Principal,
-        // Role.
-        rule = ignoreCycle(rule, "auth", "mcp");
-
-        // fleetd #131 step 2: PrimaryRegistry is used by loops in msg; move it, or put
-        // an interface between msg and mcp. Evidence: msg/ReplyPushLoop.java:5 and
-        // msg/LeadHeartbeatLoop.java:5 import mcp.PrimaryRegistry; mcp/FleetMcp.java:15-18
-        // imports msg.LeadChannel, LeadMessage, MessageService, Rendezvous.
-        rule = ignoreCycle(rule, "mcp", "msg");
-
-        // fleetd #131 -- found while implementing this test, NOT one of the ticket's
-        // original three; it names its own follow-up step before removal. Evidence:
-        // inject/CompletionResolver.java:4-5, inject/Injector.java:6 and
-        // inject/TurnListener.java:3 import msg.Rendezvous / msg.TurnToken;
-        // msg/MessageService.java:6 imports inject.Injector.
-        rule = ignoreCycle(rule, "inject", "msg");
-
-        // fleetd #131 -- same as above, its own follow-up. Evidence:
-        // metrics/FleetMetrics.java:3 imports msg.ReplyInbox; msg/MessageService.java:7-8,
-        // msg/LeadHeartbeatLoop.java:6-7 and msg/ReplyPushLoop.java:6-7 import
-        // metrics.FleetMetrics / metrics.Metrics.
-        rule = ignoreCycle(rule, "metrics", "msg");
-
-        // fleetd #131 -- same as above, its own follow-up. Evidence:
-        // session/SessionManager.java:7 imports msg.TurnToken;
-        // msg/LeadHeartbeatLoop.java:8 imports session.MemberSession.
-        rule = ignoreCycle(rule, "msg", "session");
-
+        for (String edge : BASELINE_EDGES) {
+            String[] originAndTarget = edge.split(" -> ");
+            rule = rule.ignoreDependency(originAndTarget[0], originAndTarget[1]);
+        }
         rule.check(classes);
     }
 
     /**
-     * Accepts today's known cycle between two top-level packages, and nothing else.
-     * Ignoring both directions removes exactly this pair from cycle detection; every
-     * other dependency -- including any new one added later, between these same two
-     * packages or any other pair -- is still checked.
+     * Fails with the exact offending edge when the live code and {@link #BASELINE_EDGES}
+     * disagree: a dependency crossing a baselined package pair that is not in the baseline,
+     * or a baseline entry whose dependency no longer exists.
      */
-    private static SliceRule ignoreCycle(SliceRule rule, String packageA, String packageB) {
-        return rule
-                .ignoreDependency(residesIn(packageA), residesIn(packageB))
-                .ignoreDependency(residesIn(packageB), residesIn(packageA));
+    private static void checkBaselineMatchesTodaysEdges(JavaClasses classes) {
+        Set<String> baselinedPackagePairs = new TreeSet<>();
+        for (String edge : BASELINE_EDGES) {
+            String[] originAndTarget = edge.split(" -> ");
+            baselinedPackagePairs.add(unorderedPair(
+                    topLevelPackageOf(originAndTarget[0]), topLevelPackageOf(originAndTarget[1])));
+        }
+
+        Set<String> liveEdgesInBaselinedPairs = new TreeSet<>();
+        for (JavaClass javaClass : classes) {
+            for (Dependency dependency : javaClass.getDirectDependenciesFromSelf()) {
+                JavaClass origin = dependency.getOriginClass();
+                JavaClass target = dependency.getTargetClass();
+                String originPackage = topLevelPackageOf(origin.getFullName());
+                String targetPackage = topLevelPackageOf(target.getFullName());
+                if (originPackage.isEmpty() || targetPackage.isEmpty() || originPackage.equals(targetPackage)) {
+                    continue;
+                }
+                if (baselinedPackagePairs.contains(unorderedPair(originPackage, targetPackage))) {
+                    liveEdgesInBaselinedPairs.add(origin.getFullName() + " -> " + target.getFullName());
+                }
+            }
+        }
+
+        List<String> problems = new ArrayList<>();
+        for (String liveEdge : liveEdgesInBaselinedPairs) {
+            if (!BASELINE_EDGES.contains(liveEdge)) {
+                String[] originAndTarget = liveEdge.split(" -> ");
+                problems.add("new dependency not in the baseline: " + liveEdge
+                        + " (packages " + topLevelPackageOf(originAndTarget[0])
+                        + " -> " + topLevelPackageOf(originAndTarget[1]) + ")");
+            }
+        }
+        for (String baselineEdge : BASELINE_EDGES) {
+            if (!liveEdgesInBaselinedPairs.contains(baselineEdge)) {
+                String[] originAndTarget = baselineEdge.split(" -> ");
+                problems.add("stale baseline entry, no such dependency exists: " + baselineEdge
+                        + " (packages " + topLevelPackageOf(originAndTarget[0])
+                        + " -> " + topLevelPackageOf(originAndTarget[1]) + ")");
+            }
+        }
+
+        if (!problems.isEmpty()) {
+            fail("PackageCyclesTest baseline is out of date:\n  " + String.join("\n  ", problems));
+        }
     }
 
-    private static DescribedPredicate<JavaClass> residesIn(String topLevelPackage) {
-        return Predicates.resideInAPackage("dev.ltms.fleet." + topLevelPackage + "..");
+    private static String unorderedPair(String packageA, String packageB) {
+        return packageA.compareTo(packageB) <= 0 ? packageA + "|" + packageB : packageB + "|" + packageA;
+    }
+
+    private static String topLevelPackageOf(String fullyQualifiedClassName) {
+        if (!fullyQualifiedClassName.startsWith(ROOT_PACKAGE)) {
+            return "";
+        }
+        String rest = fullyQualifiedClassName.substring(ROOT_PACKAGE.length());
+        int dot = rest.indexOf('.');
+        return dot < 0 ? "" : rest.substring(0, dot);
     }
 }

fleetd/src/test/java/dev/ltms/fleet/auth/AuthzTest.java

@@ -14,6 +14,8 @@ class AuthzTest {
     private static final Principal ANON = Principal.anonymous();
     private static final Principal ARCH_DESIGN = Principal.architect("lead-designer", "term_design", 400);
     private static final Principal ARCH_OTHER = Principal.architect("reviewer", "term_review", 500);
+    private static final Principal COLLABORATOR = Principal.collaborator("ops", "term_collab", 600);
+    private static final Principal OBSERVER = Principal.observer("term_observer", 700);
 
     @Test
     void anonymousIsAuthorizedForNothing() {
@@ -29,6 +31,22 @@ class AuthzTest {
         assertTrue(Authz.isUnauthenticated(null));
     }
 
+    /**
+     * {@code MessageService.answer}'s turn-ownership check treats a caller with no terminal as
+     * matching a turn recorded for the unnamed primary. That rule only stays safe because an
+     * unauthenticated caller — whose terminal is also {@code null} — never reaches {@code answer}
+     * at all: {@link #anonymousIsAuthorizedForNothing} already covers every action including
+     * {@code ANSWER}, but this test names the exact coupling so a future change to either side
+     * cannot drift without turning this test red.
+     */
+    @Test
+    void anAnonymousCallerIsRefusedAnswerSoItCanNeverBeMistakenForTheUnnamedPrimary() {
+        assertFalse(Authz.permits(ANON, ANSWER, null),
+                "an anonymous caller, whose terminal is also null, must never reach answer() — the "
+                        + "turn-ownership check's null-terminal match for the unnamed primary owner "
+                        + "relies on this gate refusing it first");
+    }
+
     @Test
     void orchestrationBelongsToThePrimaryAlone() {
         for (Authz.Action a : new Authz.Action[]{SPAWN, STOP, SEND, DRAIN}) {
@@ -38,6 +56,37 @@ class AuthzTest {
         }
     }
 
+    /**
+     * {@code fleet_send} is three call shapes behind one action name until {@code
+     * FleetMcp#sendAction} picks one: a plain local {@link Authz.Action#SEND}, the {@code coordId}
+     * route ({@link Authz.Action#COORD_SEND}), and the {@code turnId} answer form ({@link
+     * Authz.Action#ANSWER}). All three carry the same grant as the undivided action did — a worker
+     * is excluded from every one, exactly as it was excluded from the one combined action before.
+     */
+    @Test
+    void theThreeSendShapesCarryTheSameGrantAsTheOldUndividedAction() {
+        for (Authz.Action a : new Authz.Action[]{SEND, COORD_SEND, ANSWER}) {
+            assertTrue(Authz.permits(PRIMARY, a, "term_a"), "the primary may " + a);
+            assertTrue(Authz.permits(ARCH_DESIGN, a, "term_a"), "an architect may " + a);
+            assertFalse(Authz.permits(WORKER_A, a, "term_a"),
+                    "a worker performing " + a + " would be escalating into the orchestrator role");
+            assertFalse(Authz.permits(ANON, a, "term_a"));
+        }
+    }
+
+    /**
+     * {@code fleet_poll{ticket}} and {@code fleet_status} are {@link Authz.Action#TASK_READ}, split
+     * out of the roster-only {@link Authz.Action#READ} (fleetd #678). The grant is unchanged from
+     * what the undivided {@code READ} action gave every one of these callers.
+     */
+    @Test
+    void taskReadCarriesTheSameGrantReadDidBeforeTheSplit() {
+        assertTrue(Authz.permits(PRIMARY, TASK_READ, null));
+        assertTrue(Authz.permits(WORKER_A, TASK_READ, null));
+        assertTrue(Authz.permits(ARCH_DESIGN, TASK_READ, null));
+        assertFalse(Authz.permits(ANON, TASK_READ, null));
+    }
+
     @Test
     void aWorkerMayReplyAndAskOnlyAsItself() {
         assertTrue(Authz.permits(WORKER_A, REPLY, "term_a"));
@@ -135,4 +184,232 @@ class AuthzTest {
         assertFalse(Authz.isUnauthenticated(WORKER_A));
         assertFalse(Authz.isUnauthenticated(PRIMARY));
     }
+
+    // ── the collaborator matrix ─────────────────────────────────────────────────────────────────
+
+    /**
+     * {@code SEND} for a collaborator is the one grant that is conditional rather than fixed:
+     * flipping only the classifier's answer for the target flips only this outcome.
+     */
+    @Test
+    void aCollaboratorMaySendOnlyWhenTheClassifierAcceptsTheTarget() {
+        assertTrue(Authz.permits(COLLABORATOR, SEND, "term_lead", target -> true),
+                "the classifier accepting the target must grant SEND");
+        assertFalse(Authz.permits(COLLABORATOR, SEND, "term_lead", target -> false),
+                "the classifier refusing the target must deny SEND");
+        assertFalse(Authz.permits(COLLABORATOR, SEND, "term_lead"),
+                "the real production classifier recognises no terminal yet, so SEND is refused today");
+    }
+
+    /**
+     * Control for the test above: every other action's result for a collaborator does not move
+     * when the classifier does. Only {@code SEND} is wired to it.
+     */
+    @Test
+    void theClassifierMovesOnlySendForACollaborator() {
+        for (Authz.Action a : Authz.Action.values()) {
+            if (a == SEND) {
+                continue;
+            }
+            assertEquals(
+                    Authz.permits(COLLABORATOR, a, "term_lead"),
+                    Authz.permits(COLLABORATOR, a, "term_lead", target -> true),
+                    a + " must not depend on the classifier at all");
+        }
+    }
+
+    @Test
+    void aCollaboratorMayReadAndScrapeMetrics() {
+        assertTrue(Authz.permits(COLLABORATOR, READ, null));
+        assertTrue(Authz.permits(COLLABORATOR, METRICS, null));
+    }
+
+    @Test
+    void aCollaboratorMayReplyAndAskOnlyAsItsOwnPane() {
+        assertTrue(Authz.permits(COLLABORATOR, REPLY, "term_collab"),
+                "its own pane is its own");
+        assertTrue(Authz.permits(COLLABORATOR, ASK, "term_collab"));
+
+        assertFalse(Authz.permits(COLLABORATOR, REPLY, "term_design"),
+                "a collaborator must not reply on another pane");
+        assertFalse(Authz.permits(COLLABORATOR, REPLY, null),
+                "an absent target must not pass the own-session rule");
+    }
+
+    /**
+     * Every action denied to a collaborator, asserted denied even when the SEND classifier would
+     * accept any target — proving none of these is gated on that classifier. {@code TASK_READ} is
+     * included here with no ticket-ownership classifier supplied, so it falls back to the
+     * fail-closed default; its conditional grant for a ticket the collaborator actually created is
+     * tested separately below.
+     */
+    @Test
+    void aCollaboratorIsDeniedLifecycleCoordinationAndTicketPolling() {
+        for (Authz.Action a : new Authz.Action[]{SPAWN, STOP, DRAIN, HANDOVER, ANSWER, COORD_SEND,
+                COORD_READ, TASK_READ}) {
+            assertFalse(Authz.permits(COLLABORATOR, a, "term_lead", target -> true),
+                    "a collaborator must not " + a + " even when the classifier accepts every target");
+        }
+    }
+
+    @Test
+    void aCollaboratorIsNotCountedAsPrimaryWorkerOrArchitect() {
+        assertFalse(COLLABORATOR.isPrimary());
+        assertFalse(COLLABORATOR.isWorker());
+        assertFalse(COLLABORATOR.isArchitect());
+        assertTrue(COLLABORATOR.isCollaborator());
+    }
+
+    /**
+     * A collaborator is never spawned, so it must not be enrolled in the presence map as an
+     * available member. Control: both a worker and an architect — which ARE spawned — still are.
+     */
+    @Test
+    void isSpawnedMemberIsFalseForACollaboratorButTrueForAWorkerAndAnArchitect() {
+        assertFalse(COLLABORATOR.isSpawnedMember());
+        assertTrue(WORKER_A.isSpawnedMember());
+        assertTrue(ARCH_DESIGN.isSpawnedMember());
+    }
+
+    // ── the observer matrix ─────────────────────────────────────────────────────────────────────
+
+    @Test
+    void anObserverMayReadAndScrapeMetrics() {
+        assertTrue(Authz.permits(OBSERVER, READ, null));
+        assertTrue(Authz.permits(OBSERVER, METRICS, null));
+    }
+
+    @Test
+    void anObserverMayReplyAndAskOnlyAsItsOwnPane() {
+        assertTrue(Authz.permits(OBSERVER, REPLY, "term_observer"), "its own pane is its own");
+        assertTrue(Authz.permits(OBSERVER, ASK, "term_observer"));
+
+        assertFalse(Authz.permits(OBSERVER, REPLY, "term_design"),
+                "an observer must not reply on another pane");
+        assertFalse(Authz.permits(OBSERVER, REPLY, null),
+                "an absent target must not pass the own-session rule");
+    }
+
+    /**
+     * Every action beyond READ/METRICS/REPLY/ASK/SEND, asserted denied for an observer with no
+     * ticket-ownership classifier supplied — so {@code TASK_READ} falls back to the fail-closed
+     * default here, an unconfigured pane reading a ticket it did not create. Its conditional grant
+     * for a ticket the observer actually created is tested separately below. {@code SEND} is
+     * excluded here and given its own matrix below, since — unlike every action in this loop — its
+     * grant is conditional on the target, not fixed.
+     */
+    @Test
+    void anObserverIsDeniedEverythingBeyondReadMetricsReplyAskAndSend() {
+        for (Authz.Action a : Authz.Action.values()) {
+            if (a == READ || a == METRICS || a == REPLY || a == ASK || a == SEND) {
+                continue;
+            }
+            assertFalse(Authz.permits(OBSERVER, a, "term_observer", target -> true),
+                    "an observer must not " + a + " even when the classifier accepts every target");
+        }
+    }
+
+    @Test
+    void anObserverIsNotCountedAsAnyOtherRole() {
+        assertFalse(OBSERVER.isPrimary());
+        assertFalse(OBSERVER.isWorker());
+        assertFalse(OBSERVER.isArchitect());
+        assertFalse(OBSERVER.isCollaborator());
+        assertFalse(OBSERVER.isSpawnedMember());
+        assertTrue(OBSERVER.isObserver());
+    }
+
+    // ── the observer SEND matrix ────────────────────────────────────────────────────────────────
+
+    /**
+     * {@code SEND} for an observer is the one grant that is conditional rather than fixed, exactly
+     * like a collaborator's: flipping only the observer-target classifier's answer flips only this
+     * outcome.
+     */
+    @Test
+    void anObserverMaySendOnlyWhenTheClassifierAcceptsTheTargetAsAnObserver() {
+        assertTrue(Authz.permits(OBSERVER, SEND, "term_other_observer",
+                        Authz.NO_KNOWN_LEAD_OR_COLLABORATOR, target -> true),
+                "the classifier accepting the target as an observer must grant SEND");
+        assertFalse(Authz.permits(OBSERVER, SEND, "term_other_observer",
+                        Authz.NO_KNOWN_LEAD_OR_COLLABORATOR, target -> false),
+                "the classifier refusing the target must deny SEND");
+        assertFalse(Authz.permits(OBSERVER, SEND, "term_other_observer"),
+                "the real production classifier recognises no terminal as an observer target yet, "
+                        + "so SEND is refused by the three- and four-argument convenience forms");
+    }
+
+    /**
+     * Control for the test above: every other action's result for an observer does not move when
+     * the observer-target classifier does. Only {@code SEND} is wired to it.
+     */
+    @Test
+    void theObserverTargetClassifierMovesOnlySendForAnObserver() {
+        for (Authz.Action a : Authz.Action.values()) {
+            if (a == SEND) {
+                continue;
+            }
+            assertEquals(
+                    Authz.permits(OBSERVER, a, "term_observer"),
+                    Authz.permits(OBSERVER, a, "term_observer",
+                            Authz.NO_KNOWN_LEAD_OR_COLLABORATOR, target -> true),
+                    a + " must not depend on the observer-target classifier at all");
+        }
+    }
+
+    /**
+     * An observer's {@code SEND} is gated on a different classifier than a collaborator's: the
+     * collaborator classifier accepting every target must not itself grant an observer's SEND.
+     */
+    @Test
+    void anObserversSendDoesNotMoveOnTheCollaboratorClassifier() {
+        assertFalse(Authz.permits(OBSERVER, SEND, "term_lead", target -> true),
+                "an observer's SEND must consult the observer-target classifier, never the "
+                        + "lead-or-collaborator one");
+    }
+
+    // ── the TASK_READ ticket-ownership matrix ──────────────────────────────────────────────────────
+
+    /**
+     * {@code TASK_READ} for an observer is conditional on ticket ownership, exactly like
+     * {@code SEND} is conditional on the target: flipping only the ownership classifier's answer
+     * flips only this outcome. Paired with a negative control — the same observer, the same ticket,
+     * a classifier that reports a different creator — so a test that forgot to vary the classifier
+     * at all could not pass by accident.
+     */
+    @Test
+    void anObserverMayTaskReadATicketOnlyWhenItCreatedIt() {
+        assertTrue(Authz.permits(OBSERVER, TASK_READ, "task-1", Authz.NO_KNOWN_LEAD_OR_COLLABORATOR,
+                        Authz.NO_KNOWN_OBSERVER_TARGET, ticket -> true),
+                "a ticket the observer created must be readable");
+        assertFalse(Authz.permits(OBSERVER, TASK_READ, "task-1", Authz.NO_KNOWN_LEAD_OR_COLLABORATOR,
+                        Authz.NO_KNOWN_OBSERVER_TARGET, ticket -> false),
+                "a ticket created by someone else must still be refused to the same observer");
+        assertFalse(Authz.permits(OBSERVER, TASK_READ, "task-1"),
+                "the default classifier owns no ticket, so the three-argument form still refuses");
+    }
+
+    /** As above, for a collaborator — the grant is the same mechanism, not an observer special case. */
+    @Test
+    void aCollaboratorMayTaskReadATicketOnlyWhenItCreatedIt() {
+        assertTrue(Authz.permits(COLLABORATOR, TASK_READ, "task-1", Authz.NO_KNOWN_LEAD_OR_COLLABORATOR,
+                        Authz.NO_KNOWN_OBSERVER_TARGET, ticket -> true),
+                "a ticket the collaborator created must be readable");
+        assertFalse(Authz.permits(COLLABORATOR, TASK_READ, "task-1", Authz.NO_KNOWN_LEAD_OR_COLLABORATOR,
+                        Authz.NO_KNOWN_OBSERVER_TARGET, ticket -> false),
+                "a ticket created by someone else must still be refused to the same collaborator");
+    }
+
+    /**
+     * Control: {@code TASK_READ} for a primary, a worker, and an architect does not move on the
+     * ticket-ownership classifier — it is already unconditional for all three.
+     */
+    @Test
+    void theTicketOwnershipClassifierNeverGatesTaskReadForPrimaryWorkerOrArchitect() {
+        for (Principal p : new Principal[]{PRIMARY, WORKER_A, ARCH_DESIGN}) {
+            assertTrue(Authz.permits(p, TASK_READ, "task-1", Authz.NO_KNOWN_LEAD_OR_COLLABORATOR,
+                            Authz.NO_KNOWN_OBSERVER_TARGET, Authz.NO_CALLER_OWNS_TICKET),
+                    p.describe() + " must read a ticket unconditionally, even when the classifier owns nothing");
+        }
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/auth/CallerResolverTest.java

@@ -1,13 +1,25 @@
 package dev.ltms.fleet.auth;
 
+import dev.ltms.fleet.guard.SubscriptionGuard;
+import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.FakeHerdr;
 import dev.ltms.fleet.herdr.PaneLocator;
+import dev.ltms.fleet.herdr.WorkspaceControl;
 import dev.ltms.fleet.mcp.ConnectionIdentity;
 import dev.ltms.fleet.config.FleetConfig;
+import dev.ltms.fleet.member.ClaudeCodeLauncher;
 import dev.ltms.fleet.peer.MemberRole;
+import dev.ltms.fleet.session.MemberSession;
+import dev.ltms.fleet.session.SessionManager;
+import dev.ltms.fleet.session.WorktreeRequest;
+import dev.ltms.fleet.session.Worktrees;
 import org.junit.jupiter.api.Test;
 
+import java.util.List;
 import java.util.Map;
+import java.util.Optional;
+import java.util.Set;
+import java.util.concurrent.atomic.AtomicReference;
 
 import static org.junit.jupiter.api.Assertions.*;
 
@@ -45,17 +57,23 @@ class CallerResolverTest {
         return members;
     }
 
+    /**
+     * With no roster wired up at all (the simple constructor), a loopback pane that owns a herdr
+     * pane but is not recognised as a live spawned member lands on the {@link Role#OBSERVER} floor
+     * — unforgeable and never token-gated, exactly like a worker's own identity, because it comes
+     * from the same connection-derived pane mapping.
+     */
     @Test
-    void aLoopbackWorkerPaneResolvesToWorkerRegardlessOfAuthMode() {
+    void aLoopbackPaneWithNoLiveRosterResolvesToObserverRegardlessOfAuthMode() {
         Principal underTrust = new CallerResolver(workerIdentity()).resolve("127.0.0.1", 42, null);
         Principal underToken = new CallerResolver(workerIdentity(), true, "s3cret")
                 .resolve("127.0.0.1", 42, null);
 
-        assertEquals(Role.WORKER, underTrust.role());
+        assertEquals(Role.OBSERVER, underTrust.role());
         assertEquals("term_a", underTrust.terminal());
-        assertEquals(Role.WORKER, underToken.role(),
-                "worker identity is unforgeable and must never be token-gated — otherwise enabling "
-                        + "auth would lock the whole fleet out of fleet_reply");
+        assertEquals(Role.OBSERVER, underToken.role(),
+                "the floor is unforgeable and must never be token-gated — otherwise enabling auth "
+                        + "would lock every unconfigured pane out of even READ");
         assertEquals("term_a", underToken.terminal());
     }
 
@@ -79,20 +97,20 @@ class CallerResolverTest {
     }
 
     @Test
-    void otherPanesRemainWorkersWhenAPinIsSet() {
+    void otherPanesRemainAtTheFloorWhenAPinIsSet() {
         Principal p = CallerResolver.pinnedTo(workerIdentity(), false, null, "term_someone_else")
                 .resolve("127.0.0.1", 42, null);
 
-        assertEquals(Role.WORKER, p.role());
+        assertEquals(Role.OBSERVER, p.role());
         assertEquals("term_a", p.terminal());
     }
 
     /** The pin is optional config, so an absent or whitespace one must change nothing at all. */
     @Test
-    void aBlankPinLeavesWorkerResolutionUntouched() {
-        assertEquals(Role.WORKER,
+    void aBlankPinLeavesFloorResolutionUntouched() {
+        assertEquals(Role.OBSERVER,
                 CallerResolver.pinnedTo(workerIdentity(), false, null, "  ").resolve("127.0.0.1", 42, null).role());
-        assertEquals(Role.WORKER,
+        assertEquals(Role.OBSERVER,
                 CallerResolver.pinnedTo(workerIdentity(), false, null, null).resolve("127.0.0.1", 42, null).role());
     }
 
@@ -213,13 +231,13 @@ class CallerResolverTest {
      * mid-scan teardown into a refusal — the real match is still found and resolves as a worker.
      */
     @Test
-    void aHerdrErrorOnANonOwningPaneStillResolvesTheRealWorker() {
+    void aHerdrErrorOnANonOwningPaneStillResolvesTheRealPane() {
         FakeHerdr vanishedElsewhere = new FakeHerdr().processInfoFailsForPane("w2:p9", "pane_not_found");
         ConnectionIdentity id = new ConnectionIdentity(new PaneLocator(vanishedElsewhere), _ -> FakeHerdr.WORKER_PID);
 
         Principal p = new CallerResolver(id).resolve("127.0.0.1", 55555, null);
 
-        assertEquals(Role.WORKER, p.role());
+        assertEquals(Role.OBSERVER, p.role());
         assertEquals("term_a", p.terminal());
     }
 
@@ -263,12 +281,12 @@ class CallerResolverTest {
     }
 
     @Test
-    void aPaneAbsentFromTheRegistryIsStillAWorker() {
+    void aPaneAbsentFromTheRegistryFallsToTheObserverFloor() {
         Principal p = new CallerResolver(workerIdentity(), false, null,
                 Map.of("term_elsewhere", "gpt-sol-5.6"))
                 .resolve("127.0.0.1", 42, null);
 
-        assertEquals(Role.WORKER, p.role());
+        assertEquals(Role.OBSERVER, p.role());
         assertEquals("term_a", p.terminal());
         assertNull(p.name());
     }
@@ -294,16 +312,16 @@ class CallerResolverTest {
     }
 
     @Test
-    void anEmptyRegistryLeavesEveryPaneAWorker() {
+    void anEmptyRegistryLeavesEveryPaneAtTheObserverFloor() {
         Map<String, String> noLeads = null;
-        assertEquals(Role.WORKER,
+        assertEquals(Role.OBSERVER,
                 new CallerResolver(workerIdentity(), false, null, Map.of())
                         .resolve("127.0.0.1", 42, null).role());
-        assertEquals(Role.WORKER,
+        assertEquals(Role.OBSERVER,
                 new CallerResolver(workerIdentity(), false, null, noLeads)
                         .resolve("127.0.0.1", 42, null).role());
-        // CB-531: and the same for the live-registry form, whose supplier may also be absent.
-        assertEquals(Role.WORKER,
+        // And the same for the live-registry form, whose supplier may also be absent.
+        assertEquals(Role.OBSERVER,
                 CallerResolver.withLeads(workerIdentity(), false, null, null)
                         .resolve("127.0.0.1", 42, null).role());
     }
@@ -376,7 +394,7 @@ class CallerResolverTest {
         Map<String, String> live = new java.util.HashMap<>();
         CallerResolver r = CallerResolver.withLeads(workerIdentity(), false, null, () -> live);
 
-        assertEquals(Role.WORKER, r.resolve("127.0.0.1", 42, null).role());
+        assertEquals(Role.OBSERVER, r.resolve("127.0.0.1", 42, null).role());
 
         live.put("term_a", "gpt-sol-5.6"); // the scanner sees a newly-labelled tab
 
@@ -393,7 +411,7 @@ class CallerResolverTest {
 
         mutable.put("term_a", "sneaky");
 
-        assertEquals(Role.WORKER, r.resolve("127.0.0.1", 42, null).role());
+        assertEquals(Role.OBSERVER, r.resolve("127.0.0.1", 42, null).role());
     }
 
     // ── CB-548: architect slots ─────────────────────────────────────────────────────────────────
@@ -423,13 +441,13 @@ class CallerResolverTest {
     }
 
     @Test
-    void anUnboundPaneStillResolvesAsAWorker() {
+    void anUnboundPaneResolvesToTheObserverFloor() {
         MemberRegistry members = new MemberRegistry(new FleetConfig.Fleet(Map.of(),
                 Map.of("lead-designer", new FleetConfig.Slot("sonnet")), Map.of(), Map.of(), null));
         Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
                 Map::of, members).resolve("127.0.0.1", 42, null);
 
-        assertEquals(Role.WORKER, p.role());
+        assertEquals(Role.OBSERVER, p.role());
         assertNull(p.name());
     }
 
@@ -455,12 +473,11 @@ class CallerResolverTest {
         CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
                 Map::of, members);
 
-        assertEquals(Role.WORKER, r.resolve("127.0.0.1", 42, null).role());
+        assertEquals(Role.OBSERVER, r.resolve("127.0.0.1", 42, null).role());
 
         assertTrue(members.bind("architect:lead-designer", "term_a")); // the later lifecycle binds the slot
 
         assertEquals(Role.ARCHITECT, r.resolve("127.0.0.1", 42, null).role());
-        assertEquals("architect:lead-designer", r.members().get("term_a"));
     }
 
     @Test
@@ -483,12 +500,13 @@ class CallerResolverTest {
     }
 
     @Test
-    void aBoundNonArchitectSlotStillResolvesAsAWorker() {
+    void aBoundNonArchitectSlotResolvesToTheObserverFloorNotArchitect() {
         Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
                 Map::of, boundMembers("dev:builder", MemberRole.DEV))
                 .resolve("127.0.0.1", 42, null);
 
-        assertEquals(Role.WORKER, p.role(), "a dev binding must never grant architect rights");
+        assertEquals(Role.OBSERVER, p.role(), "a dev binding must never grant architect rights, "
+                + "and this construction path wires no roster to recognise it as the live dev it is");
     }
 
     @Test
@@ -499,17 +517,17 @@ class CallerResolverTest {
         assertThrows(IllegalArgumentException.class, () -> new CallerResolver(id, true, "  "));
     }
     @Test
-    void aWorkerOnAnyLoopbackSourceAddressIsStillAWorkerNotThePrimary() {
-        // fleetd #305: the escalation. ConnectionIdentity used to accept only 127.0.0.1, so a
-        // worker connecting from 127.0.0.2 resolved to no terminal, and this resolver's own
-        // (wider) loopback check then made it the PRIMARY — granting spawn, stop, send and drain.
-        // Measured on the Linux fleet host: binding a source of 127.0.0.2 succeeds there, so the
-        // path is real and not theoretical.
+    void aPaneOnAnyLoopbackSourceAddressIsStillAtTheFloorNotThePrimary() {
+        // fleetd #305: the escalation this guards against. ConnectionIdentity used to accept only
+        // 127.0.0.1, so a pane connecting from 127.0.0.2 resolved to no terminal, and this
+        // resolver's own (wider) loopback check then made it the PRIMARY — granting spawn, stop,
+        // send and drain. Measured on the Linux fleet host: binding a source of 127.0.0.2 succeeds
+        // there, so the path is real and not theoretical.
         CallerResolver r = new CallerResolver(workerIdentity(), false, null);
         for (String src : new String[]{"127.0.0.1", "127.0.0.2", "127.1.2.3", "::ffff:127.0.0.2"}) {
             Principal p = r.resolve(src, 55555, null);
-            assertEquals(Role.WORKER, p.role(), "a worker must stay a worker from source " + src);
-            assertEquals("term_a", p.terminal(), "worker terminal from source " + src);
+            assertEquals(Role.OBSERVER, p.role(), "the pane must stay off PRIMARY from source " + src);
+            assertEquals("term_a", p.terminal(), "pane terminal from source " + src);
         }
     }
 
@@ -523,4 +541,443 @@ class CallerResolverTest {
         }
     }
 
+    // ── fleetd #669 Unit D: a live spawned member outranks every tab map ────────────────────────
+
+    /**
+     * Criterion 1: a terminal present in BOTH the spawned-member roster AND the lead tab map
+     * resolves as its member role, not as a lead — the roster is checked first, consulting no tab
+     * map at all when it matches.
+     */
+    @Test
+    void aSpawnedMemberWinsOverALeadTabForTheSamePane() {
+        Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
+                        () -> Map.of("term_a", "opus-5.0"), new MemberRegistry(null),
+                        t -> "term_a".equals(t) ? MemberRole.DEV : null, Map::of)
+                .resolve("127.0.0.1", 42, null);
+
+        assertEquals(Role.WORKER, p.role(),
+                "a live spawned member's own identity must win over a tab map naming the same pane a lead");
+        assertEquals("term_a", p.terminal());
+    }
+
+    /**
+     * fleetd #702: between {@link SessionManager#release} removing the registry entry and the
+     * pane actually stopping, a resolve for that terminal must still see the live member and
+     * never fall through to a tab map — wired through the real {@link SessionManager}, not a
+     * hand-rolled stand-in for its {@code spawnedMemberRole}.
+     *
+     * <p>Reuses the ticket's own test idea: an injected {@code hasUncommitted} resolves the
+     * releasing terminal from inside {@code release}'s window — a real call landing inside the
+     * window, so no sleep and no race.
+     *
+     * <p>The property asserted is "no tab map is consulted", never "the same role is returned".
+     * The mandatory control is the lead tab map: it names this exact terminal, and the same
+     * resolve taken <em>outside</em> the window (before release runs) must still return the lead
+     * role — without that control, the in-window assertion would also pass on an empty map and
+     * prove nothing.
+     */
+    @Test
+    void aPaneMidTeardownResolvesAsItsOwnRoleConsultingNoTabMap() {
+        FakeHerdr herdr = new FakeHerdr().pinNextStarts(1, "term_a", "w2:p7");
+        FleetConfig.Profile cfg = new FleetConfig.Profile(
+                "ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN",
+                List.of("ccs", "ltms-local"), "tab", "fleetd-workers",
+                "worker: {profile} #{n}", null, null, null);
+        ClaudeCodeLauncher workers = new ClaudeCodeLauncher(new AgentControl(herdr), new WorkspaceControl(herdr),
+                new SubscriptionGuard(Set.of("gx00.gw")), Map.of(cfg.profile(), cfg), cfg.profile(), _ -> null);
+
+        AtomicReference<Principal> duringWindow = new AtomicReference<>();
+        AtomicReference<CallerResolver> resolverRef = new AtomicReference<>();
+        Worktrees worktrees = new Worktrees() {
+            @Override
+            public String add(String repoRoot, String branch, String baseRef) {
+                return "/wt/" + branch.replace('/', '_');
+            }
+
+            @Override
+            public void remove(String repoRoot, String worktreePath) {
+            }
+
+            @Override
+            public void deleteBranch(String repoRoot, String branch) {
+            }
+
+            @Override
+            public boolean hasUncommitted(String worktreePath) {
+                // Runs from INSIDE release()'s git-status shell-out: the registry entry is
+                // already gone, but the pane has not stopped yet.
+                duringWindow.set(resolverRef.get().resolve("127.0.0.1", 42, null));
+                return false;
+            }
+
+            @Override
+            public void overlayParity(String repoRoot, String worktreePath, List<String> overlay) {
+            }
+
+            @Override
+            public String repoRoot(String cwd) {
+                return "/repo";
+            }
+
+            @Override
+            public Optional<String> snapshot(String worktreePath, String branch, String message) {
+                return Optional.empty();
+            }
+
+            @Override
+            public WipRefStats wipRefs(String repoRoot) {
+                return new WipRefStats(0, 0L);
+            }
+
+            @Override
+            public int pruneWipRefs(String repoRoot, long minAgeMillis) {
+                return 0;
+            }
+
+            @Override
+            public void shareWithGroup(String repoRoot, String worktreePath) {
+            }
+        };
+        SessionManager sessions = new SessionManager(workers, worktrees);
+
+        // The lead tab map names "term_a" before anything is ever spawned onto it — the control
+        // this test needs. Built up front so the SAME resolver answers every resolve() call below.
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> FakeHerdr.WORKER_PID);
+        CallerResolver resolver = CallerResolver.withLeadsAndMembers(identity, false, null,
+                () -> Map.of("term_a", "the-lead"), null, sessions::spawnedMemberRole, Map::of);
+        resolverRef.set(resolver);
+
+        Principal before = resolver.resolve("127.0.0.1", 42, null);
+        assertEquals(Role.PRIMARY, before.role(),
+                "control: with no live or releasing member on this terminal, the lead tab map must "
+                        + "win — this is what proves the in-window assertion below is not passing on "
+                        + "an empty map");
+        assertEquals("the-lead", before.name());
+
+        MemberSession s = sessions.acquire("ltms-local", null, "/caller/proj", null,
+                new WorktreeRequest("fleetd-702", null));
+        assertEquals("term_a", s.terminalId(), "sanity: the spawn resolved to the pinned pane");
+
+        sessions.release(s.paneId());
+
+        assertNotNull(duringWindow.get(), "the dirty check must have run and captured a resolve");
+        assertEquals(Role.WORKER, duringWindow.get().role(),
+                "inside the window the pane must resolve as its own live-member role, consulting no "
+                        + "tab map — a lead tab naming the same terminal must not win");
+        assertEquals("term_a", duringWindow.get().terminal());
+    }
+
+    /**
+     * {@link SessionManager#releaseRemoved} unbinds the architect slot before the git-status
+     * shell-out that opens the teardown window, so a resolve landing inside that window must see
+     * the slot already unbound and resolve {@link Role#WORKER} — never {@link Role#ARCHITECT},
+     * and never by checking role equality against the live session, which would hold even if the
+     * unbind ran too late.
+     *
+     * <p>The control is the same resolve taken outside the window, while the slot is still bound,
+     * which must return {@link Role#ARCHITECT} — without it this test would also pass against a
+     * slot that was never bound, and prove nothing.
+     */
+    @Test
+    void aReleasingArchitectIsDemotedToWorkerInsideTheTeardownWindow() {
+        FakeHerdr herdr = new FakeHerdr().pinNextStarts(1, "term_a", "w2:p7");
+        FleetConfig.Profile cfg = new FleetConfig.Profile(
+                "ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN",
+                List.of("ccs", "ltms-local"), "tab", "fleetd-workers",
+                "worker: {profile} #{n}", null, null, null);
+        ClaudeCodeLauncher workers = new ClaudeCodeLauncher(new AgentControl(herdr), new WorkspaceControl(herdr),
+                new SubscriptionGuard(Set.of("gx00.gw")), Map.of(cfg.profile(), cfg), cfg.profile(), _ -> null);
+
+        AtomicReference<Principal> duringWindow = new AtomicReference<>();
+        AtomicReference<CallerResolver> resolverRef = new AtomicReference<>();
+        Worktrees worktrees = new Worktrees() {
+            @Override
+            public String add(String repoRoot, String branch, String baseRef) {
+                return "/wt/" + branch.replace('/', '_');
+            }
+
+            @Override
+            public void remove(String repoRoot, String worktreePath) {
+            }
+
+            @Override
+            public void deleteBranch(String repoRoot, String branch) {
+            }
+
+            @Override
+            public boolean hasUncommitted(String worktreePath) {
+                // Runs from INSIDE release()'s git-status shell-out: the architect slot is already
+                // unbound by this point, but the pane has not stopped yet.
+                duringWindow.set(resolverRef.get().resolve("127.0.0.1", 42, null));
+                return false;
+            }
+
+            @Override
+            public void overlayParity(String repoRoot, String worktreePath, List<String> overlay) {
+            }
+
+            @Override
+            public String repoRoot(String cwd) {
+                return "/repo";
+            }
+
+            @Override
+            public Optional<String> snapshot(String worktreePath, String branch, String message) {
+                return Optional.empty();
+            }
+
+            @Override
+            public WipRefStats wipRefs(String repoRoot) {
+                return new WipRefStats(0, 0L);
+            }
+
+            @Override
+            public int pruneWipRefs(String repoRoot, long minAgeMillis) {
+                return 0;
+            }
+
+            @Override
+            public void shareWithGroup(String repoRoot, String worktreePath) {
+            }
+        };
+        SessionManager sessions = new SessionManager(workers, worktrees);
+
+        MemberRegistry members = new MemberRegistry(new FleetConfig.Fleet(Map.of(),
+                Map.of("lead-designer", new FleetConfig.Slot("ltms-local")), Map.of(), Map.of(), null));
+        sessions.setMemberLifecycle(members);
+
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> FakeHerdr.WORKER_PID);
+        CallerResolver resolver = CallerResolver.withLeadsAndMembers(identity, false, null,
+                Map::of, members, sessions::spawnedMemberRole, Map::of);
+        resolverRef.set(resolver);
+
+        MemberSession s = sessions.acquire("ltms-local", MemberRole.ARCHITECT, null, "/caller/proj", null,
+                new WorktreeRequest("fleetd-702d", null));
+        assertEquals("term_a", s.terminalId(), "sanity: the spawn resolved to the pinned pane");
+        assertEquals(MemberRole.ARCHITECT, s.role(), "sanity: the slot bind succeeded");
+
+        Principal before = resolver.resolve("127.0.0.1", 42, null);
+        assertEquals(Role.ARCHITECT, before.role(),
+                "control: with the slot still bound, the pane must resolve as an architect — this "
+                        + "is what proves the in-window assertion below is not passing against a "
+                        + "slot that was never bound");
+        assertEquals("lead-designer", before.name());
+
+        sessions.release(s.paneId());
+
+        assertNotNull(duringWindow.get(), "the dirty check must have run and captured a resolve");
+        assertEquals(Role.WORKER, duringWindow.get().role(),
+                "inside the window the architect slot is already unbound, so the result must be "
+                        + "WORKER — asserting role-equality with the live session here would tempt "
+                        + "moving the unbind earlier or later, which would be wrong either way");
+        assertEquals("term_a", duringWindow.get().terminal());
+    }
+
+    /** A spawned architect in the roster resolves ARCHITECT, carrying its bound slot's name. */
+    @Test
+    void aSpawnedArchitectInTheRosterResolvesArchitectWithItsSlotName() {
+        Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                        boundMembers("architect:lead-designer", MemberRole.ARCHITECT),
+                        t -> "term_a".equals(t) ? MemberRole.ARCHITECT : null, Map::of)
+                .resolve("127.0.0.1", 42, null);
+
+        assertEquals(Role.ARCHITECT, p.role());
+        assertEquals("lead-designer", p.name());
+        assertEquals("term_a", p.terminal());
+    }
+
+    /**
+     * fleetd #424 regression: the roster only answers THAT a pane is a live spawned member; config
+     * still decides WHAT that member's slot grants. A slot revoked after the bind must still demote
+     * the session on its very next request, exactly as it would for a pane with no roster entry at
+     * all — the roster's own ARCHITECT role must never be granted on its word alone.
+     *
+     * <p>{@code bind} refuses an unconfigured slot, so the revoked state can only be reached by
+     * binding while the slot is configured and then swapping the config out from under it, the way
+     * a live reload does.
+     */
+    @Test
+    void aRevokedArchitectSlotDemotesALiveSpawnedArchitectToWorker() {
+        FleetConfig.Fleet configured = new FleetConfig.Fleet(Map.of(),
+                Map.of("lead-designer", new FleetConfig.Slot("sonnet")), Map.of(), Map.of(), null);
+        java.util.concurrent.atomic.AtomicReference<FleetConfig.Fleet> live =
+                new java.util.concurrent.atomic.AtomicReference<>(configured);
+        MemberRegistry members = MemberRegistry.live(live::get);
+        assertTrue(members.bind("architect:lead-designer", "term_a"));
+
+        live.set(new FleetConfig.Fleet(Map.of(), Map.of(), Map.of(), Map.of(), null)); // slot revoked
+
+        // Setup controls: the slot is really gone from config, but the occupancy is still there —
+        // otherwise this test would pass for the wrong reason.
+        assertNull(members.roleForSlot("architect:lead-designer"), "setup control: the slot must be gone from config");
+        assertEquals("architect:lead-designer", members.snapshot().get("term_a"),
+                "setup control: the binding itself must still be there");
+
+        Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                        members, t -> "term_a".equals(t) ? MemberRole.ARCHITECT : null, Map::of)
+                .resolve("127.0.0.1", 42, null);
+
+        assertEquals(Role.WORKER, p.role(),
+                "a revoked slot must demote a live spawned architect on its very next request");
+    }
+
+    /** Criterion 3: a configured collaborator tab that is not a spawned member resolves COLLABORATOR. */
+    @Test
+    void aConfiguredCollaboratorTabResolvesToCollaboratorCarryingItsName() {
+        Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                        new MemberRegistry(null), t -> null, () -> Map.of("term_a", "ops"))
+                .resolve("127.0.0.1", 42, null);
+
+        assertEquals(Role.COLLABORATOR, p.role());
+        assertEquals("ops", p.name());
+        assertEquals("term_a", p.terminal());
+    }
+
+    /** Regression: an empty collaborator registry leaves every pane at the unconfigured-pane floor. */
+    @Test
+    void anEmptyCollaboratorRegistryLeavesEveryPaneAtTheObserverFloor() {
+        Principal p = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                        new MemberRegistry(null), t -> null, Map::of)
+                .resolve("127.0.0.1", 42, null);
+
+        assertEquals(Role.OBSERVER, p.role());
+        assertNull(p.name());
+    }
+
+    @Test
+    void describeNamesTheCollaborator() {
+        assertEquals("collaborator:ops", Principal.collaborator("ops", "term_a", 1).describe());
+    }
+
+    // ── fleetd #669 Unit D: knownLeadOrCollaborator() reads the same maps resolve() does ───────────
+
+    @Test
+    void knownLeadOrCollaboratorIsTrueForALeadTerminal() {
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
+                () -> Map.of("term_lead", "opus-5.0"), new MemberRegistry(null), t -> null, Map::of);
+
+        assertTrue(r.knownLeadOrCollaborator().test("term_lead"));
+        assertFalse(r.knownLeadOrCollaborator().test("term_other"));
+    }
+
+    @Test
+    void knownLeadOrCollaboratorIsTrueForACollaboratorTerminal() {
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                new MemberRegistry(null), t -> null, () -> Map.of("term_collab", "ops"));
+
+        assertTrue(r.knownLeadOrCollaborator().test("term_collab"));
+        assertFalse(r.knownLeadOrCollaborator().test("term_other"));
+    }
+
+    // ── fleetd #705: narrowing the unconfigured-pane floor to OBSERVER ──────────────────────────
+
+    /**
+     * The case this ticket exists for: a pane the resolver cannot place as a live spawned member,
+     * a lead, a bound architect slot, or a configured collaborator must land on the narrow
+     * {@link Role#OBSERVER} floor, never the {@link Role#WORKER} the old fallback granted.
+     *
+     * <p>The second assertion is the control the ticket requires: a terminal the roster DOES
+     * recognise as a live spawned member must still resolve its own role. Without it, this test
+     * would also pass if the fix accidentally turned every caller into an observer.
+     */
+    @Test
+    void anUnconfiguredPaneResolvesObserverButARegisteredMemberStillResolvesItsOwnRole() {
+        Principal unconfigured = new CallerResolver(workerIdentity()).resolve("127.0.0.1", 42, null);
+        assertEquals(Role.OBSERVER, unconfigured.role(),
+                "a pane matching none of the configured or live-roster roles must fall to the "
+                        + "floor, not WORKER");
+        assertEquals("term_a", unconfigured.terminal());
+
+        Principal registered = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
+                        Map::of, new MemberRegistry(null),
+                        t -> "term_a".equals(t) ? MemberRole.DEV : null, Map::of)
+                .resolve("127.0.0.1", 42, null);
+        assertEquals(Role.WORKER, registered.role(),
+                "control: a live spawned member must keep resolving its own role, never the "
+                        + "unconfigured-pane floor");
+    }
+
+    @Test
+    void describeNamesTheObserverByItsPane() {
+        assertEquals("observer:term_a", Principal.observer("term_a", 1).describe());
+    }
+
+    @Test
+    void knownLeadOrCollaboratorIsFalseForASpawnedMembersTerminal() {
+        // The exact scenario a collaborator's SEND must never reach: a live spawned member's own
+        // terminal, which is neither a configured lead nor a configured collaborator.
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                new MemberRegistry(null), t -> "term_a".equals(t) ? MemberRole.DEV : null, Map::of);
+
+        assertFalse(r.knownLeadOrCollaborator().test("term_a"));
+    }
+
+    // ── fleetd #743: sendableObserverTarget() reads the same maps and functions resolve() does ────
+
+    /**
+     * A terminal this resolver recognises as none of the privileged roles is exactly the one
+     * {@code resolve} would itself hand back {@link Role#OBSERVER} for.
+     */
+    @Test
+    void sendableObserverTargetIsTrueForATerminalKnownAsNoOtherRole() {
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
+                () -> Map.of("term_lead", "opus-5.0"), new MemberRegistry(null),
+                t -> "term_worker".equals(t) ? MemberRole.DEV : null,
+                () -> Map.of("term_collab", "ops"));
+
+        assertTrue(r.sendableObserverTarget().test("term_other"));
+    }
+
+    @Test
+    void sendableObserverTargetIsFalseForALeadTerminal() {
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null,
+                () -> Map.of("term_lead", "opus-5.0"), new MemberRegistry(null), t -> null, Map::of);
+
+        assertFalse(r.sendableObserverTarget().test("term_lead"),
+                "a lead's own terminal must never be a sendable observer target");
+    }
+
+    @Test
+    void sendableObserverTargetIsFalseForACollaboratorTerminal() {
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                new MemberRegistry(null), t -> null, () -> Map.of("term_collab", "ops"));
+
+        assertFalse(r.sendableObserverTarget().test("term_collab"),
+                "a collaborator's own terminal must never be a sendable observer target");
+    }
+
+    @Test
+    void sendableObserverTargetIsFalseForALiveSpawnedMembersTerminal() {
+        // Covers both a worker and an architect: spawnedMemberRole.apply(target) is non-null for
+        // either, and resolve() never falls through to OBSERVER once it is.
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                new MemberRegistry(null),
+                t -> switch (t) {
+                    case "term_worker" -> MemberRole.DEV;
+                    case "term_architect" -> MemberRole.ARCHITECT;
+                    default -> null;
+                }, Map::of);
+
+        assertFalse(r.sendableObserverTarget().test("term_worker"));
+        assertFalse(r.sendableObserverTarget().test("term_architect"));
+    }
+
+    @Test
+    void sendableObserverTargetIsFalseForABoundArchitectSlotWithNoLiveMember() {
+        // The edge case resolve() itself carries: a terminal bound to a configured architect slot
+        // but with no live spawned-member session yet.
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                boundMembers("architect:lead-designer", MemberRole.ARCHITECT), t -> null, Map::of);
+
+        assertFalse(r.sendableObserverTarget().test("term_a"));
+    }
+
+    @Test
+    void sendableObserverTargetIsFalseForANullTarget() {
+        CallerResolver r = CallerResolver.withLeadsAndMembers(workerIdentity(), false, null, Map::of,
+                new MemberRegistry(null), t -> null, Map::of);
+
+        assertFalse(r.sendableObserverTarget().test(null));
+    }
+
 }

fleetd/src/test/java/dev/ltms/fleet/auth/MemberRegistryLiveTest.java

@@ -295,9 +295,10 @@ class MemberRegistryLiveTest {
         assertTrue(out.applied(), "the reload must actually take effect: " + out.summary());
 
         Principal after = resolver.resolve("127.0.0.1", 42, null);
-        assertEquals(Role.WORKER, after.role(),
-                "removing the slot from config must demote the bound session to worker on its "
-                        + "NEXT request — this is the ticket's whole point");
+        assertEquals(Role.OBSERVER, after.role(),
+                "removing the slot from config must demote the bound session on its NEXT request — "
+                        + "this harness wires no live roster for term_a, so the demotion lands on "
+                        + "the unconfigured-pane floor");
         assertEquals("term_a", after.terminal(), "same pane, same terminal — only the role changed");
     }
 

fleetd/src/test/java/dev/ltms/fleet/auth/PrincipalTest.java

@@ -0,0 +1,31 @@
+package dev.ltms.fleet.auth;
+
+import org.junit.jupiter.api.Test;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertNotEquals;
+import static org.junit.jupiter.api.Assertions.assertNull;
+
+class PrincipalTest {
+
+    @Test
+    void ownerKeyCoversEveryRole() {
+        assertEquals("leader:opus", Principal.leader("opus", "term_lead", 1).ownerKey());
+        assertNull(Principal.primary(2).ownerKey());
+        assertEquals("worker:term_worker", Principal.worker("term_worker", 3).ownerKey());
+        assertEquals("architect:term_arch", Principal.architect("opus", "term_arch", 4).ownerKey());
+        assertEquals("collaborator:ops", Principal.collaborator("ops", "term_collab", 5).ownerKey());
+        assertEquals("observer:term_observer", Principal.observer("term_observer", 6).ownerKey());
+        assertEquals("anonymous", Principal.anonymous().ownerKey());
+    }
+
+    @Test
+    void rolePrefixesKeepLeadAndArchitectKeysDistinct() {
+        String lead = Principal.leader("opus", "term_lead", 1).ownerKey();
+        String architect = Principal.architect("design", "opus", 2).ownerKey();
+
+        assertEquals("leader:opus", lead);
+        assertEquals("architect:opus", architect);
+        assertNotEquals(lead, architect);
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/config/ConfigRefTest.java

@@ -107,6 +107,8 @@ class ConfigRefTest {
 
         assertTrue(out.applied());
         assertTrue(out.deferred().isEmpty());
+        assertTrue(out.split().stream().noneMatch(s -> s.contains("fleet.collaborators")),
+                out.split().toString());
         assertEquals("new charter", ref.get().fleet().charterFor(
                 dev.ltms.fleet.peer.MemberRole.ARCHITECT));
     }
@@ -786,14 +788,100 @@ class ConfigRefTest {
         assertTrue(out.split().getFirst().startsWith("fleet:"), out.split().toString());
         assertTrue(out.split().getFirst().contains("restart"), out.split().toString());
         assertTrue(out.split().getFirst().contains("live"), out.split().toString());
+        assertTrue(out.split().stream().noneMatch(s -> s.contains("fleet.collaborators")),
+                out.split().toString());
         assertTrue(out.summary().contains("partially live"), out.summary());
         // The snapshot still carries the new value — the LeadTabScanner's identity map and
         // LeadLauncher's auto-launch are what wait for a restart; a reload rebuilds neither.
         assertEquals("lead: opus-b", ref.get().fleet().leaders().get("opus").tab());
     }
 
+    @Test
+    void addingAFleetCollaboratorIsReportedAsSplit(@TempDir Path dir) throws Exception {
+        assertCollaboratorChangeIsReported(dir, """
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collaborator: alex"
+                """, "add a collaborator");
+    }
+
+    @Test
+    void removingAFleetCollaboratorIsReportedAsSplit(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("fleetd.yaml");
+        Files.writeString(f, yaml("""
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collaborator: alex"
+                """));
+        ConfigRef ref = refFor(f);
+
+        Files.writeString(f, yaml("fleet: {}\n"));
+        assertCollaboratorSplit(ref.reload(), "remove a collaborator");
+    }
+
+    @Test
+    void changingAFleetCollaboratorTabIsReportedAsSplit(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("fleetd.yaml");
+        Files.writeString(f, yaml("""
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collaborator: alex-a"
+                """));
+        ConfigRef ref = refFor(f);
+
+        Files.writeString(f, yaml("""
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collaborator: alex-b"
+                """));
+        assertCollaboratorSplit(ref.reload(), "change a collaborator tab");
+    }
+
+    @Test
+    void unchangedFleetCollaboratorsProduceNoSplitReport(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("fleetd.yaml");
+        String config = yaml("""
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collaborator: alex"
+                """);
+        Files.writeString(f, config);
+        ConfigRef ref = refFor(f);
+
+        Files.writeString(f, config);
+        ConfigRef.Outcome out = ref.reload();
+
+        assertTrue(out.applied());
+        assertTrue(out.split().isEmpty(), out.split().toString());
+    }
+
+    private static void assertCollaboratorChangeIsReported(Path dir, String changed, String action)
+            throws Exception {
+        Path f = dir.resolve("fleetd.yaml");
+        Files.writeString(f, yaml("fleet: {}\n"));
+        ConfigRef ref = refFor(f);
+
+        Files.writeString(f, yaml(changed));
+        assertCollaboratorSplit(ref.reload(), action);
+    }
+
+    private static void assertCollaboratorSplit(ConfigRef.Outcome out, String action) {
+        assertTrue(out.applied(), action);
+        assertTrue(out.deferred().isEmpty(), out.deferred().toString());
+        assertEquals(1, out.split().size(), out.split().toString());
+        String report = out.split().getFirst();
+        assertTrue(report.startsWith("fleet: fleet.collaborators"), report);
+        assertTrue(report.contains("read once at startup"), report);
+        assertTrue(report.contains("restart"), report);
+    }
+
     /**
-     * fleetd #333: {@code fleet.leaders} is the ONLY frozen part of {@code fleet:}. A reload that
+     * fleetd #333: {@code fleet.leaders} is a frozen part of {@code fleet:}. A reload that
      * changes {@code tabLabel} (or charters, or a role pool) without touching {@code fleet.leaders}
      * must stay fully hot with nothing reported — proving {@link ConfigRef#changedSplitKeys}
      * compares {@code fleet.leaders} specifically rather than the whole {@code Fleet} record, which

fleetd/src/test/java/dev/ltms/fleet/config/FleetConfigTest.java

@@ -537,12 +537,13 @@ class FleetConfigTest {
     }
 
     /**
-     * CB-579: {@code tab} is the only field a lead's identity depends on now, so it is required
-     * whether the entry is creatable or recognise-only — without it the entry can never be found.
+     * A lead's tab label is a fixed constant, not a per-entry field, so an entry with no {@code
+     * tab:} is the normal case — it is still found by that constant label in its own {@code
+     * workspace}, not refused as useless.
      */
     @Test
-    void aLeadWithNoTabRefusesToStart(@TempDir Path dir) throws Exception {
-        Path f = dir.resolve("useless-lead.yaml");
+    void aLeadWithNoTabIsAcceptedAndFoundByTheFixedLabel(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("no-tab-lead.yaml");
         Files.writeString(f, """
                 bind:
                   port: 8080
@@ -553,9 +554,9 @@ class FleetConfigTest {
                 """);
         FleetConfig cfg = FleetConfig.load(f);
 
-        IllegalStateException e = assertThrows(IllegalStateException.class, cfg::validateMembers);
-        assertTrue(e.getMessage().contains("ghost"), "the message must name the useless entry");
-        assertTrue(e.getMessage().contains("tab:"), "the message must say what is missing");
+        assertDoesNotThrow(cfg::validateMembers);
+        assertEquals(List.of(FleetConfig.Leader.LEAD_TAB_LABEL),
+                cfg.fleet().leaders().get("ghost").acceptedLabels());
     }
 
     /**
@@ -676,12 +677,8 @@ class FleetConfigTest {
         assertEquals(5, hb.quietNudgeCap());
     }
 
-    /**
-     * The hazard the guard exists for: fleetd writes worker tab labels and reads lead tab labels.
-     * Overlap the two and every worker it spawns is read back as a lead.
-     */
     @Test
-    void aLeadPrefixThatAProfileTabLabelOverrideAlsoMatchesRefusesToStart(@TempDir Path dir)
+    void aProfileTabLabelOverrideMatchingALeadTabRefusesToStart(@TempDir Path dir)
             throws Exception {
         Path f = dir.resolve("collide.yaml");
         Files.writeString(f, """
@@ -689,32 +686,33 @@ class FleetConfigTest {
                   port: 8080
                 profiles:
                   gx10:
-                    tabLabel: "lead: {profile} #{n}"
+                    tabLabel: "alpha"
                 fleet:
                   leaders:
                     opus:
-                      tab: "lead: opus"
-                      tabPrefix: "lead:"
+                      tab: "alpha"
                 """);
         FleetConfig cfg = FleetConfig.load(f);
 
         IllegalStateException e =
                 assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
         assertTrue(e.getMessage().contains("gx10"), "the message must name the offending profile");
+        assertTrue(e.getMessage().contains("alpha"), "the message must name the offending label");
     }
 
-    /** A bad fleet-wide template promotes every member, not one profile — so it is checked too. */
     @Test
-    void aFleetTabLabelThatMatchesALeadPrefixRefusesToStart(@TempDir Path dir) throws Exception {
+    void aFleetTabLabelTemplateThatCanRenderAsALeadTabRefusesToStart(@TempDir Path dir) throws Exception {
         Path f = dir.resolve("collide-template.yaml");
         Files.writeString(f, """
                 bind:
                   port: 8080
+                profiles:
+                  pha: {}
                 fleet:
-                  tabLabel: "lead: {role} {profile}"
+                  tabLabel: "al{profile}"
                   leaders:
                     opus:
-                      tab: "lead: opus"
+                      tab: "alpha"
                 """);
         FleetConfig cfg = FleetConfig.load(f);
 
@@ -723,12 +721,28 @@ class FleetConfigTest {
         assertTrue(e.getMessage().contains("fleet.tabLabel"));
     }
 
-    /**
-     * The point of making role the label's first field: {@code {role}} comes from a closed enum, so
-     * a generated label cannot begin with {@code "lead:"} however the fleet is configured.
-     */
     @Test
-    void theDefaultTabLabelCannotCollideWithTheDefaultLeadPrefix(@TempDir Path dir) throws Exception {
+    void anExactFleetTabLabelCollisionRefusesToStart(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("exact-tab-collision.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  tabLabel: "alpha"
+                  leaders:
+                    alpha:
+                      tab: "alpha"
+                """);
+
+        IllegalStateException e = assertThrows(IllegalStateException.class,
+                () -> FleetConfig.load(f).validateAll());
+        assertTrue(e.getMessage().contains("fleet.tabLabel"),
+                "the message must name the offending label");
+        assertTrue(e.getMessage().contains("alpha"), "the message must name the colliding lead tab");
+    }
+
+    @Test
+    void aFleetTabLabelTemplateThatCannotRenderAsALeadTabIsAllowed(@TempDir Path dir) throws Exception {
         Path f = dir.resolve("ok.yaml");
         Files.writeString(f, """
                 bind:
@@ -737,17 +751,13 @@ class FleetConfigTest {
                   gx10:
                     baseUrl: http://gx00.gw:8000
                 fleet:
+                  tabLabel: "worker-{profile}"
                   leaders:
                     opus:
-                      tab: "lead: opus"
+                      tab: "alpha"
                 """);
 
-        assertDoesNotThrow(() -> FleetConfig.load(f).validateLeadTabPrefixes());
-        for (MemberRole role : MemberRole.values()) {
-            assertFalse(FleetConfig.Fleet.DEFAULT_TAB_LABEL
-                            .replace("{role}", role.wireName()).startsWith("lead:"),
-                    "no role renders a label that reads as a lead");
-        }
+        assertDoesNotThrow(() -> FleetConfig.load(f).validateAll());
     }
 
     @Test
@@ -765,11 +775,132 @@ class FleetConfigTest {
                 "a label that collides with a convention nobody reads is not a problem");
     }
 
+    /**
+     * A lead's tab label is fixed, so two leads sharing one {@code workspace} would both resolve
+     * to the one tab named {@code lead} there — the guard must catch this independently of the
+     * member-template checks above.
+     */
+    @Test
+    void twoLeadersSharingTheSameWorkspaceRefuseToStart(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("shared-workspace.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  leaders:
+                    opus:
+                      workspace: "shared"
+                    sonnet:
+                      workspace: "shared"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("opus"), "the message must name one offending lead");
+        assertTrue(e.getMessage().contains("sonnet"), "the message must name the other offending lead");
+        assertTrue(e.getMessage().contains("shared"), "the message must name the shared workspace");
+    }
+
+    /** The workspace collision check is case-insensitive, matching how spaces are looked up. */
+    @Test
+    void twoLeadersSharingTheSameWorkspaceInDifferentCaseRefuseToStart(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("shared-workspace-case.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  leaders:
+                    opus:
+                      workspace: "Shared"
+                    sonnet:
+                      workspace: "shared"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("opus"), "the message must name one offending lead");
+        assertTrue(e.getMessage().contains("sonnet"), "the message must name the other offending lead");
+    }
+
+    /** Control for the two tests above: distinct workspaces load cleanly, with no {@code tab:} at all. */
+    @Test
+    void twoLeadersWithDistinctWorkspacesAndNoTabAreAllowed(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("distinct-workspaces.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  leaders:
+                    opus:
+                      workspace: "space-opus"
+                    sonnet:
+                      workspace: "space-sonnet"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        assertDoesNotThrow(cfg::validateLeadTabPrefixes);
+    }
+
+    /**
+     * A member tab-label template that can render exactly as the fixed lead tab label would let a
+     * member's own tab be read back as a lead — refused outright, with no lead needing to be
+     * configured at all.
+     */
+    @Test
+    void aFleetTabLabelTemplateThatCanRenderAsTheFixedLeadTabLabelRefusesToStart(@TempDir Path dir)
+            throws Exception {
+        Path f = dir.resolve("template-renders-as-lead.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  tabLabel: "lead"
+                  leaders:
+                    opus:
+                      workspace: "fleet"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("fleet.tabLabel"),
+                "the message must name the offending template");
+        assertTrue(e.getMessage().contains(FleetConfig.Leader.LEAD_TAB_LABEL),
+                "the message must name the fixed lead tab label it collides with");
+    }
+
+    /**
+     * A collaborator's {@code tab} equal to the fixed lead tab label would shadow a lead sharing
+     * that space — refused outright.
+     */
+    @Test
+    void aCollaboratorTabEqualToTheFixedLeadTabLabelRefusesToStart(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("collaborator-is-lead.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  collaborators:
+                    impostor:
+                      tab: "lead"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("impostor"), "the message must name the offending collaborator");
+        assertTrue(e.getMessage().contains(FleetConfig.Leader.LEAD_TAB_LABEL),
+                "the message must name the fixed lead tab label it collides with");
+    }
+
     // ── validatePanePlacementAgainstLeadTabs ────────────────────────────────────────────────────
 
     /**
      * The hazard this guard closes: a pane-placed member lands inside the focused tab rather than
-     * its own, so it can land inside a lead's labelled tab and be read back as that lead.
+     * its own, so it can land inside a lead's labelled tab and, while its pane carries no entry in
+     * the spawned-member roster, be read back as that lead.
      */
     @Test
     void aPanePlacedProfileWithALeadTabRefusesToStart(@TempDir Path dir) throws Exception {
@@ -792,8 +923,12 @@ class FleetConfigTest {
         assertTrue(e.getMessage().contains("gx10"), "the message must name the offending profile");
     }
 
+    /**
+     * A lead is found by its fixed tab label regardless of its own {@code tab} field, so a
+     * {@code fleet.leaders} entry with no {@code tab} must still arm the guard.
+     */
     @Test
-    void aPanePlacedProfileWithNoLeadTabIsAllowed(@TempDir Path dir) throws Exception {
+    void aPanePlacedProfileWithNoLeadTabRefusesToStart(@TempDir Path dir) throws Exception {
         Path f = dir.resolve("pane-no-tab.yaml");
         Files.writeString(f, """
                 bind:
@@ -806,9 +941,59 @@ class FleetConfigTest {
                     opus:
                       profile: gx10
                 """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e = assertThrows(IllegalStateException.class,
+                cfg::validatePanePlacementAgainstLeadTabs);
+        assertTrue(e.getMessage().contains("gx10"), "the message must name the offending profile");
+        assertTrue(e.getMessage().contains("the only fix when a lead triggered this"),
+                "the message must say placement: tab is the only fix for a lead");
+        assertFalse(e.getMessage().contains("remove the tab from every fleet.leaders"),
+                "the message must not send the operator in a circle by advising a tab: removal");
+    }
+
+    /**
+     * {@code placement:} is optional, and {@link FleetConfig.Profile}'s own compact constructor
+     * defaults an absent or blank value to {@code "tab"}, so a profile naming no placement at all
+     * is tab-placed and the guard must not fire for it.
+     */
+    @Test
+    void aProfileWithNoPlacementKeyDefaultsToTabPlacementAndIsAllowed(@TempDir Path dir)
+            throws Exception {
+        Path f = dir.resolve("no-placement-key.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                profiles:
+                  gx10: {}
+                fleet:
+                  leaders:
+                    opus:
+                      profile: gx10
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        assertTrue(cfg.profiles().get("gx10").tabPlacement(),
+                "Profile's compact constructor defaults an absent placement to \"tab\"");
+        assertDoesNotThrow(cfg::validatePanePlacementAgainstLeadTabs,
+                "a profile with no placement: key is tab-placed, not pane-placed");
+    }
+
+    /** Control: no {@code fleet.leaders} entry and no collaborator tab still starts fine. */
+    @Test
+    void aPanePlacedProfileWithAnEmptyFleetBlockIsAllowed(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("pane-empty-fleet.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                profiles:
+                  gx10:
+                    placement: pane
+                fleet: {}
+                """);
 
         assertDoesNotThrow(() -> FleetConfig.load(f).validatePanePlacementAgainstLeadTabs(),
-                "a leader with no tab feeds nothing into the scanner, so pane placement is safe");
+                "no fleet.leaders entry and no collaborator tab means pane placement is safe");
     }
 
     @Test
@@ -919,6 +1104,274 @@ class FleetConfigTest {
                 "no primary.terminal pin ⇒ nothing registered, even with fleet.leaders configured");
     }
 
+    // ── fleetd #669: the collaborators registry ────────────────────────────────────────────────
+
+    /**
+     * {@code Fleet} is {@code @JsonIgnoreProperties(ignoreUnknown = true)}, so a config naming
+     * {@code fleet.collaborators.<name>.tab} loads with no exception whether or not the key is
+     * ever read into the object model. Asserting only "no exception" would pass both before and
+     * after the real fix, so this asserts the parsed value is actually reachable from the loaded
+     * {@code FleetConfig} — the one thing a vacuous "no exception" test cannot tell apart.
+     */
+    @Test
+    void collaboratorsBlockIsActuallyParsedNotSilentlyDropped(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("collaborators.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  collaborators:
+                    reviewer-alex:
+                      tab: "collab: alex"
+                """);
+
+        FleetConfig cfg = FleetConfig.load(f);
+        assertEquals("collab: alex", cfg.fleet().collaborators().get("reviewer-alex").tab());
+    }
+
+    /** A collaborator carries no field other than {@code tab}, so a blank one is meaningless. */
+    @Test
+    void aCollaboratorWithNoTabRefusesToStart(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("useless-collaborator.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  collaborators:
+                    ghost: {}
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e = assertThrows(IllegalStateException.class, cfg::validateMembers);
+        assertTrue(e.getMessage().contains("ghost"), "the message must name the useless entry");
+        assertTrue(e.getMessage().contains("tab:"), "the message must say what is missing");
+    }
+
+    /** Control for {@link #aCollaboratorWithNoTabRefusesToStart}: a named tab loads cleanly. */
+    @Test
+    void aCollaboratorWithATabIsAllowed(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("named-collaborator.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  collaborators:
+                    reviewer-alex:
+                      tab: "collab: alex"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        assertDoesNotThrow(cfg::validateMembers);
+    }
+
+    /**
+     * fleetd #669: the fleet-wide {@code tabLabel} template can render as a collaborator tab, the
+     * same hazard {@link #aFleetTabLabelTemplateThatCanRenderAsALeadTabRefusesToStart} covers on
+     * the lead side. Drives the fleet-wide branch directly, with no profile override involved.
+     */
+    @Test
+    void aFleetTabLabelTemplateThatCanRenderAsACollaboratorTabRefusesToStart(@TempDir Path dir)
+            throws Exception {
+        Path f = dir.resolve("collide-template-collaborator.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  tabLabel: "al{profile}"
+                  collaborators:
+                    alex:
+                      tab: "alpha"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("fleet.tabLabel"));
+        assertTrue(e.getMessage().contains("alex"), "the message must name the offending collaborator");
+    }
+
+    /**
+     * A member tabLabel that can render as a configured collaborator tab is the same hazard as the
+     * lead case above — while its pane carries no entry in the spawned-member roster, a member
+     * labelled that way is read back as the collaborator.
+     */
+    @Test
+    void aProfileTabLabelOverrideMatchingACollaboratorTabRefusesToStart(@TempDir Path dir)
+            throws Exception {
+        Path f = dir.resolve("collide-collaborator.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                profiles:
+                  gx10:
+                    tabLabel: "collab-tab"
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collab-tab"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("gx10"), "the message must name the offending profile");
+        assertTrue(e.getMessage().contains("collab-tab"), "the message must name the offending label");
+    }
+
+    /** Control: a profile tabLabel that cannot render as the collaborator tab is allowed. */
+    @Test
+    void aProfileTabLabelThatCannotRenderAsACollaboratorTabIsAllowed(@TempDir Path dir)
+            throws Exception {
+        Path f = dir.resolve("ok-collaborator.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                profiles:
+                  gx10:
+                    tabLabel: "worker-{profile}"
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collab-tab"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        assertDoesNotThrow(cfg::validateLeadTabPrefixes);
+    }
+
+    /** fleetd #669: identity is matched on a collaborator's exact tab, so two sharing one are unreachable. */
+    @Test
+    void twoCollaboratorsSharingTheSameExactTabRefusesToStart(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("shared-collaborator-tab.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "shared tab"
+                    sam:
+                      tab: "Shared Tab"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("alex"), "the message must name one offending collaborator");
+        assertTrue(e.getMessage().contains("sam"), "the message must name the other offending collaborator");
+    }
+
+    /** Control for {@link #twoCollaboratorsSharingTheSameExactTabRefusesToStart}: distinct tabs load cleanly. */
+    @Test
+    void twoCollaboratorsWithDistinctExactTabsAreAllowed(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("distinct-collaborator-tabs.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "alex tab"
+                    sam:
+                      tab: "sam tab"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        assertDoesNotThrow(cfg::validateLeadTabPrefixes);
+    }
+
+    /**
+     * fleetd #669: a collaborator tab equal to a lead tab crosses a privilege boundary — the worst
+     * of the three new collisions, since only one of the two identities is ever found.
+     */
+    @Test
+    void aCollaboratorTabEqualToALeadTabRefusesToStart(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("lead-collaborator-collision.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  leaders:
+                    opus:
+                      tab: "shared tab"
+                  collaborators:
+                    alex:
+                      tab: "Shared Tab"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, cfg::validateLeadTabPrefixes);
+        assertTrue(e.getMessage().contains("opus"), "the message must name the offending lead");
+        assertTrue(e.getMessage().contains("alex"), "the message must name the offending collaborator");
+    }
+
+    /** Control for {@link #aCollaboratorTabEqualToALeadTabRefusesToStart}: distinct tabs load cleanly. */
+    @Test
+    void aLeadAndACollaboratorWithDistinctTabsAreAllowed(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("lead-collaborator-ok.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  leaders:
+                    opus:
+                      tab: "lead tab"
+                  collaborators:
+                    alex:
+                      tab: "collab tab"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        assertDoesNotThrow(cfg::validateLeadTabPrefixes);
+    }
+
+    /**
+     * fleetd #669: a pane-placed member can land in a collaborator's labelled tab exactly as it
+     * can land in a lead's — {@code validatePanePlacementAgainstLeadTabs()} must fire even when
+     * {@code fleet.leaders} is empty, which is the early-return the brief flagged as the bug.
+     */
+    @Test
+    void aPanePlacedProfileWithACollaboratorTabRefusesToStartEvenWithNoLeaders(@TempDir Path dir)
+            throws Exception {
+        Path f = dir.resolve("pane-hazard-collaborator.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                profiles:
+                  gx10:
+                    placement: pane
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collab: alex"
+                """);
+        FleetConfig cfg = FleetConfig.load(f);
+
+        IllegalStateException e = assertThrows(IllegalStateException.class,
+                cfg::validatePanePlacementAgainstLeadTabs);
+        assertTrue(e.getMessage().contains("gx10"), "the message must name the offending profile");
+    }
+
+    /** Control: a pane-placed profile with no lead or collaborator tab configured is allowed. */
+    @Test
+    void aPanePlacedProfileWithNoLeaderOrCollaboratorTabIsAllowed(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("pane-no-tab-at-all.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                profiles:
+                  gx10:
+                    placement: pane
+                fleet:
+                  collaborators:
+                    alex: {}
+                """);
+
+        assertDoesNotThrow(() -> FleetConfig.load(f).validatePanePlacementAgainstLeadTabs(),
+                "a collaborator with no tab feeds nothing into the scanner, so pane placement is safe");
+    }
+
     // ── CB-548: the architects registry ────────────────────────────────────────────────────────
 
     @Test
@@ -1191,6 +1644,60 @@ class FleetConfigTest {
         assertEquals(Set.of("sonnet"), cfg.fleet().pool(MemberRole.REVIEWER).keySet());
     }
 
+    /**
+     * A duplicated name in {@code fleet.collaborators} is refused at parse time, like any other
+     * {@code fleet:} pool. See {@link #duplicateSlotNamesInOnePoolAreRejectedAtParseTime}.
+     */
+    @Test
+    void duplicateCollaboratorNamesAreRejectedAtParseTime(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("collaborator-dup.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                fleet:
+                  collaborators:
+                    alex:
+                      tab: "collab: alex"
+                    alex:
+                      tab: "collab: alex, second"
+                """);
+
+        IllegalStateException e =
+                assertThrows(IllegalStateException.class, () -> FleetConfig.load(f));
+        assertTrue(e.getMessage().contains("alex"),
+                "the refusal names the duplicated entry, was: " + e.getMessage());
+        assertTrue(e.getMessage().contains("fleet.collaborators"),
+                "the refusal names the pool the duplicate is in, was: " + e.getMessage());
+    }
+
+    /**
+     * Control for {@link #duplicateCollaboratorNamesAreRejectedAtParseTime}: the same name reused
+     * across the collaborators registry and a member role pool is the role × profile matrix doing
+     * its job in the other pool, not a mistake — only a repeat within one pool loses an entry.
+     */
+    @Test
+    void theSameNameInCollaboratorsAndAnotherPoolIsNotADuplicate(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("collaborator-cross-pool.yaml");
+        Files.writeString(f, """
+                bind:
+                  port: 8080
+                profiles:
+                  sonnet:
+                    baseUrl: http://gx10.gw:8000
+                fleet:
+                  architects:
+                    alex:
+                      profile: sonnet
+                  collaborators:
+                    alex:
+                      tab: "collab: alex"
+                """);
+
+        FleetConfig cfg = assertDoesNotThrow(() -> FleetConfig.load(f));
+        assertEquals(Set.of("alex"), cfg.fleet().pool(MemberRole.ARCHITECT).keySet());
+        assertEquals(Set.of("alex"), cfg.fleet().collaborators().keySet());
+    }
+
     @Test
     void duplicateKeysOutsideTheFleetPoolsAreUnaffected(@TempDir Path dir) throws Exception {
         // The duplicate check is scoped to the fleet pools — a duplicate elsewhere is not this

fleetd/src/test/java/dev/ltms/fleet/config/FleetConfigValidateAllTest.java

@@ -17,63 +17,21 @@ import static org.junit.jupiter.api.Assertions.assertThrows;
 import static org.junit.jupiter.api.Assertions.assertTrue;
 
 /**
- * The gap this class exists to close: mutation testing on the fleetd ticket "central allow-list
- * of usable models" found that although {@link FleetConfig#validateModels()}'s own logic was well
- * pinned, nothing proved either real caller ({@code Fleetd.main} and {@link ConfigRef#reload()})
- * still invoked it — deleting the call site left the full suite green (1478/0/0/0). A follow-up
- * measurement (same technique — remove one call site, run the suite, not read the code) found the
- * SAME gap for all five of {@link FleetConfig}'s other validators at startup, and for four of the
- * six inside {@link ConfigRef#reload()}. This is a class of gap, not one line's mistake: every one
- * of those thirteen tests called the validator itself directly, never the real caller that was
- * supposed to.
+ * Tests the reflective validator sweep and {@link FleetConfig#validateAll()} reachability.
  *
- * <p>The fix replaces the six individual {@code cfg.validateXxx()} calls at each of the two real
- * call sites with one {@link FleetConfig#validateAll()}, which reaches every validator by
- * reflection rather than by a hand-maintained list of names. A hand-maintained list of six names
- * would have exactly the defect it replaces: the seventh validator someone adds next month has no
- * reason to be added to it, and nothing would say so. This class proves TWO separate claims, and
- * keeps them separate on purpose:
- *
- * <ol>
- *   <li>{@link #theSweepMechanismIsGenericNotHardcodedToFleetConfigsSixNames()} and its neighbours
- *       prove the reflective sweep itself ({@link FleetConfig#invokeAllValidators}) is a general
- *       mechanism — it runs whatever public, no-arg, void {@code validateXxx()} methods a class
- *       happens to declare today, including a class with more of them than {@link FleetConfig}
- *       has right now. This is the proof that a future, real seventh validator on {@link
- *       FleetConfig} would be swept automatically, without needing to add a real (unwanted)
- *       seventh validator just to exercise the claim.</li>
- *   <li>{@link #validateAllReachesEveryOneOfTodaysRealValidators()} proves {@link
- *       FleetConfig#validateAll()} itself is wired to that same generic mechanism and genuinely
- *       reaches every one of today's real validators — reusing the exact minimal failing
- *       configurations {@code FleetConfigTest} already established for each one directly, plus a
- *       dedicated fixture for {@link FleetConfig#validateLeadRollover()}, which no other test
- *       drives through {@code validateAll()} — so a single call to {@code validateAll()} is shown
- *       to reproduce every one of those failures.</li>
- * </ol>
- *
- * <p>Together with the direct-{@code Fleetd.main}-invocation tests in {@code
- * FleetdStartupValidationTest} (which prove the real startup call site still calls {@code
- * validateAll()}) and the {@code ConfigRefTest} reload tests (which prove the same for {@link
- * ConfigRef#reload()}), removing {@code cfg.validateAll();} from either real call site now fails
- * a test in this module.
- *
- * <p><b>What is NOT pinned, measured rather than assumed.</b> Reverting {@link
- * FleetConfig#validateAll()} to a hardcoded list of today's method calls leaves the whole
- * suite green. Nothing ties {@code validateAll()} to
- * the generic sweep — claim 1 proves {@link FleetConfig#invokeAllValidators} is generic, and claim
- * 2 proves {@code validateAll()} reaches today's validators, and a hardcoded list satisfies both. So the
- * reflective sweep is a convenience, not the guarantee. The guarantee is {@link
- * #fleetConfigDeclaresExactlyTheseValidatorsToday()}: it fails the moment any validator is added
- * or removed, which forces whoever changes the set to look at this file.
+ * <p>{@link #theSweepRunsEveryValidateMethodOnAnUnrelatedClass()} and its neighbours
+ * prove that {@link FleetConfig#invokeAllValidators} runs each public, no-arg, void
+ * {@code validateXxx()} method on its target. {@link #fleetConfigDeclaresExactlyTheseValidatorsToday()}
+ * is the canary for the validator set. {@link #validateAllReachesEveryOneOfTodaysRealValidators()}
+ * is the reachability check for that set.
  */
 class FleetConfigValidateAllTest {
 
-    // ── Claim 1: the reflective sweep is a general mechanism, not six names in disguise ──────────
+    // ── Claim 1: the reflective sweep is a general mechanism ─────────────────────────────────────
 
     /**
-     * A throwaway fixture class, unrelated to {@link FleetConfig} in every way except shape: three
-     * public, no-arg, void methods named {@code validateXxx}. Proves the sweep works on ANY class
-     * with this shape, not on something special-cased to {@link FleetConfig}.
+     * Fixture with public, no-arg, void methods named {@code validateXxx}. It proves the sweep uses
+     * the target's method shape rather than special handling for {@link FleetConfig}.
      */
     static class ThreeValidators {
         final List<String> ran = new ArrayList<>();
@@ -92,7 +50,7 @@ class FleetConfigValidateAllTest {
     }
 
     @Test
-    void theSweepMechanismIsGenericNotHardcodedToFleetConfigsSixNames() {
+    void theSweepRunsEveryValidateMethodOnAnUnrelatedClass() {
         ThreeValidators target = new ThreeValidators();
         FleetConfig.invokeAllValidators(target);
         assertEquals(List.of("validateAlpha", "validateBeta", "validateGamma"), target.ran,
@@ -102,12 +60,8 @@ class FleetConfigValidateAllTest {
     }
 
     /**
-     * The core of the "self-maintaining" requirement: the exact same class shape as {@link
-     * ThreeValidators}, plus one more method — standing in for "a developer adds a validator next
-     * month". Nothing about the sweep changes to pick it up; the new method is invoked purely
-     * because it exists and matches the shape. This is what makes adding a seventh real validator
-     * to {@link FleetConfig} safe without touching {@link FleetConfig#validateAll()} or either
-     * call site — there is no "wire it in" step left to forget.
+     * Fixture with an added valid method. It proves the sweep reaches a method because it matches
+     * the validator shape.
      */
     static class FourValidators {
         final List<String> ran = new ArrayList<>();
@@ -135,7 +89,7 @@ class FleetConfigValidateAllTest {
         FleetConfig.invokeAllValidators(target);
         assertEquals(List.of("validateAlpha", "validateBeta", "validateDelta", "validateGamma"),
                 sorted(target.ran),
-                "the fourth method must be reached automatically — proving a class can grow the "
+                "the added method must be reached automatically — proving a class can grow the "
                         + "set of things it validates with no change to the sweep itself");
     }
 
@@ -297,16 +251,16 @@ class FleetConfigValidateAllTest {
                   port: 8765
                 """, "auth.mode: token");
 
-        // validateLeadTabPrefixes: a fleet-wide tabLabel that starts with a lead's own tabPrefix.
+        // validateLeadTabPrefixes: a fleet-wide tabLabel that equals a lead tab.
         assertValidateAllRefuses(dir, "lead-tab-prefixes.yaml", """
                 bind:
                   host: 127.0.0.1
                   port: 8765
                 fleet:
-                  tabLabel: "lead: {role} {profile}"
+                  tabLabel: "alpha"
                   leaders:
                     opus:
-                      tab: "lead: opus"
+                      tab: "alpha"
                 """, "fleet.tabLabel");
 
         // validateSubscriptionProfiles: subscription: true with env: reseating ANTHROPIC_BASE_URL.

fleetd/src/test/java/dev/ltms/fleet/config/FleetConfigWithDefaultsPreservesEveryComponentTest.java

@@ -103,7 +103,7 @@ class FleetConfigWithDefaultsPreservesEveryComponentTest {
         // comment there), same as broker/primary/leadHeartbeat/... above — a real, non-null value
         // here proves it, rather than leaving it null and proving nothing.
         v.put("leadRollover", new FleetConfig.LeadRollover(
-                "/handover/guard.md", true, 3600, 20, 20, "read the handover file"));
+                "/handover/guard.md", true, 3600, 20, 45, "read the handover file"));
         assertNamesMatchComponents(v);
         return v;
     }

fleetd/src/test/java/dev/ltms/fleet/herdr/FakeHerdr.java

@@ -7,6 +7,7 @@ import java.util.ArrayList;
 import java.util.LinkedHashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.Set;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.concurrent.CopyOnWriteArrayList;
 
@@ -23,6 +24,41 @@ public final class FakeHerdr implements HerdrClient {
     /** The foreground PID of the one agent pane (term_a) in the canned {@code pane.process_info}. */
     public static final long WORKER_PID = 4242;
 
+    /**
+     * A {@code detection} region of the current Claude Code TUI, whose input box is empty: a caret
+     * line between two rules, above the footer.
+     */
+    public static final String IDLE_PROMPT_CARET = """
+            ──────────────────────── lead: opus ─
+            ❯
+            ─────────────────────────────────────
+              lead: opus · Opus 5 (1M context) · ~/LTMS/claude-bridge
+              ⏵⏵ auto mode on (shift+tab to cycle) · ← 1 agent""";
+
+    /** The same region with the operator's unsubmitted line still at the caret. */
+    public static final String DRAFTED_PROMPT_CARET = """
+            ──────────────────────── lead: opus ─
+            ❯ yes, send it to lead: opus
+            ─────────────────────────────────────
+              lead: opus · Opus 5 (1M context) · ~/LTMS/claude-bridge
+              ⏵⏵ auto mode on (shift+tab to cycle) · ← 1 agent""";
+
+    /** A {@code detection} region of an older TUI, which drew a bordered box, with that box empty. */
+    public static final String IDLE_PROMPT_BOX = """
+            ⏺ done
+            ╭────────────────────────╮
+            │ >                      │
+            ╰────────────────────────╯
+              ⏵⏵ auto mode on""";
+
+    /** The older TUI's bordered box, still holding the operator's unsubmitted line. */
+    public static final String DRAFTED_PROMPT_BOX = """
+            ⏺ done
+            ╭────────────────────────╮
+            │ > fix the issue when I │
+            ╰────────────────────────╯
+              ⏵⏵ auto mode on""";
+
     private final ObjectMapper mapper = new ObjectMapper();
     /**
      * Thread-safe on purpose. Background loops — {@link dev.ltms.fleet.msg.ReplyPushLoop} and the
@@ -47,18 +83,25 @@ public final class FakeHerdr implements HerdrClient {
     private final Map<String, String> processInfoErrorCodeFor = new ConcurrentHashMap<>();
     private String tabCloseErrorCode = null;
     private final Map<String, String> tabCloseErrorCodeFor = new ConcurrentHashMap<>();
+    private String workspaceListErrorCode = null;
     private String agentSendErrorCode = null;
     private boolean noPanes = false;
     private volatile String agentStatus = "idle"; // steady-state agent.get status
     private volatile String agentType = "claude"; // detected agent kind on agent.get; null = undetected
     private volatile String agentSessionId = null; // agent_session.value on agent.get; null = omitted
     private volatile String readText = "worker transcript tail"; // canned agent.read output
+    /** Canned {@code detection}-source output, or {@code null} to serve {@link #readText} there too. */
+    private volatile String detectionText = null;
     private int pinnedStarts = 0;   // how many upcoming agent.start calls report a fixed pane
     private String pinnedStartTerminal;
     private String pinnedStartPane;
     private Runnable onAgentStart; // fires the instant agent.start is called — see onAgentStart(Runnable)
     private volatile int agentGetOkCalls = Integer.MAX_VALUE; // how many agent.get calls succeed first
     private volatile String agentGetFailCode = null; // error code every agent.get call after that reports
+    /** pane ids that {@link #paneGoneAfterClose} has opted into reporting gone — see that method. */
+    private final Set<String> paneGoneAfterCloseIds = ConcurrentHashMap.newKeySet();
+    /** pane ids a {@code pane.close} call has actually reached, for {@link #paneGoneAfterCloseIds}. */
+    private final Set<String> closedPaneIds = ConcurrentHashMap.newKeySet();
 
     public FakeHerdr healthy(boolean h) {
         this.healthy = h;
@@ -137,6 +180,12 @@ public final class FakeHerdr implements HerdrClient {
         return this;
     }
 
+    /** Make {@code workspace.list} fail with this herdr error code; every other method still succeeds. */
+    public FakeHerdr workspaceListFailsWith(String code) {
+        this.workspaceListErrorCode = code;
+        return this;
+    }
+
     /**
      * Make {@code pane.list} report no panes at all — models a second herdr daemon (CB-185) that
      * simply does not host the pane a {@link PaneLocator} is searching for.
@@ -195,12 +244,26 @@ public final class FakeHerdr implements HerdrClient {
         return this;
     }
 
-    /** The text {@code agent.read} returns (the CB-106 completion scrape). */
+    /**
+     * The text {@code agent.read} returns (the CB-106 completion scrape). It serves the
+     * {@code detection} source as well unless {@link #detectionText} overrides that one.
+     */
     public FakeHerdr readText(String text) {
         this.readText = text;
         return this;
     }
 
+    /**
+     * Override the text {@code agent.read} returns for the {@code detection} source only — the
+     * prompt/footer tail herdr uses for status detection, a different region from the transcript the
+     * other sources carry. Needed by a test whose subject reads the input box, since one
+     * {@link #readText} cannot be both a worker's transcript and a lead's empty prompt.
+     */
+    public FakeHerdr detectionText(String text) {
+        this.detectionText = text;
+        return this;
+    }
+
     /** Make delivery ({@code agent.prompt} / {@code agent.send_keys}) fail with this error code. */
     public FakeHerdr agentSendFailsWith(String code) {
         this.agentSendErrorCode = code;
@@ -220,6 +283,19 @@ public final class FakeHerdr implements HerdrClient {
         return this;
     }
 
+    /**
+     * Make {@code pane.get(paneId)} report the pane gone (a {@code pane_not_found} {@link
+     * HerdrException}, exactly as {@link WorkspaceControl#locatePane} expects to see once a pane
+     * has really disappeared) once a {@code pane.close} call for that same {@code paneId} has
+     * actually reached this fake. Every other pane, and this pane before its own close, keeps
+     * reporting the default canned {@code pane.get} response — opt-in, by pane id, so no existing
+     * test's {@code pane.get} behaviour changes.
+     */
+    public FakeHerdr paneGoneAfterClose(String paneId) {
+        paneGoneAfterCloseIds.add(paneId);
+        return this;
+    }
+
     /**
      * Run {@code hook} synchronously the instant an {@code agent.start} call reaches this fake —
      * i.e. the instant the peer PROCESS would start against a real herdr daemon. A test uses this
@@ -284,11 +360,17 @@ public final class FakeHerdr implements HerdrClient {
                 case "ping" -> mapper.readTree(
                         ("{\"type\":\"pong\",\"version\":\"%s\",\"protocol\":%d}")
                                 .formatted(pingVersion, pingProtocol));
-                case "workspace.list" -> mapper.readTree(("""
+                case "workspace.list" -> {
+                    if (workspaceListErrorCode != null) {
+                        throw new HerdrException("herdr error [" + workspaceListErrorCode + "]: workspace.list failed",
+                                workspaceListErrorCode, null);
+                    }
+                    yield mapper.readTree(("""
                         {"type":"workspace_list","workspaces":[
                           {"workspace_id":"w1","label":"dev-mgnl","focused":true,"pane_count":7,"agent_status":"unknown"},
                           {"workspace_id":"w2","label":"ltms","focused":false,"pane_count":5,"agent_status":"done"}%s]}""")
                         .formatted(extraWorkspaces.isEmpty() ? "" : "," + String.join(",", extraWorkspaces)));
+                }
                 case "agent.list" -> mapper.readTree(("""
                         {"type":"agent_list","agents":[
                           {"terminal_id":"term_a","agent":"claude","agent_status":"idle",
@@ -329,8 +411,13 @@ public final class FakeHerdr implements HerdrClient {
                           "agent_status":"%s","workspace_id":"w2","tab_id":"w2:t7","pane_id":"w2:p7"%s}}""")
                             .formatted(agentField, agentStatus, sessionField));
                 }
-                case "agent.read" -> mapper.readTree(mapper.writeValueAsString(
-                        java.util.Map.of("type", "agent_read", "read", java.util.Map.of("text", readText))));
+                case "agent.read" -> {
+                    Object source = params instanceof Map<?, ?> m ? m.get("source") : null;
+                    String text = "detection".equals(source) && detectionText != null
+                            ? detectionText : readText;
+                    yield mapper.readTree(mapper.writeValueAsString(
+                            java.util.Map.of("type", "agent_read", "read", java.util.Map.of("text", text))));
+                }
                 case "agent.start" -> {
                     if (onAgentStart != null) {
                         onAgentStart.run();
@@ -422,9 +509,18 @@ public final class FakeHerdr implements HerdrClient {
                     }
                     yield mapper.readTree("{\"type\":\"ok\"}");
                 }
-                case "pane.get" -> mapper.readTree("""
+                case "pane.get" -> {
+                    Object paneIdParam = params instanceof Map<?, ?> m ? m.get("pane_id") : null;
+                    String paneIdKey = paneIdParam == null ? null : String.valueOf(paneIdParam);
+                    if (paneIdKey != null && paneGoneAfterCloseIds.contains(paneIdKey)
+                            && closedPaneIds.contains(paneIdKey)) {
+                        throw new HerdrException("herdr error [pane_not_found]: pane.get failed",
+                                "pane_not_found", null);
+                    }
+                    yield mapper.readTree("""
                         {"type":"pane_info","pane":{"pane_id":"w9:pW","workspace_id":"w9",
                           "tab_id":"w9:t2","agent_status":"idle"}}""");
+                }
                 case "pane.list" -> noPanes
                         ? mapper.readTree("{\"type\":\"pane_list\",\"panes\":[]}")
                         : mapper.readTree("""
@@ -458,6 +554,9 @@ public final class FakeHerdr implements HerdrClient {
                         throw new HerdrException("herdr error [" + code + "]: pane.close failed",
                                 code, null);
                     }
+                    if (paneIdParam != null) {
+                        closedPaneIds.add(String.valueOf(paneIdParam));
+                    }
                     yield mapper.readTree("{\"type\":\"ok\"}");
                 }
                 default -> throw new HerdrException("fake has no canned response for " + method);

fleetd/src/test/java/dev/ltms/fleet/herdr/HerdrRouterTest.java

@@ -2,6 +2,8 @@ package dev.ltms.fleet.herdr;
 
 import org.junit.jupiter.api.Test;
 
+import java.util.Map;
+
 import static org.junit.jupiter.api.Assertions.assertSame;
 import static org.junit.jupiter.api.Assertions.assertNotSame;
 
@@ -28,4 +30,44 @@ class HerdrRouterTest {
         assertSame(router.leadAgents(), router.agentsFor("lead"));
         assertSame(router.memberAgents(), router.agentsFor("member"));
     }
+
+    /**
+     * fleetd #669 Unit E. The predicate shape here is exactly what {@code FleetdAssembly} builds:
+     * true when the terminal is a known lead OR a known collaborator. Two distinct clients are
+     * required — with one shared client {@code agentsFor} would return the same object regardless
+     * of the predicate's answer, and this assertion would pass whether or not the collaborator map
+     * was ever consulted.
+     */
+    @Test
+    void collaboratorTerminalRoutesToTheLeadDaemon() {
+        FakeHerdr lead = new FakeHerdr();
+        FakeHerdr member = new FakeHerdr();
+        Map<String, String> leads = Map.of("term_lead", "primary");
+        Map<String, String> collaborators = Map.of("term_collab", "reviewer-alex");
+        HerdrRouter router = new HerdrRouter(lead, member,
+                id -> leads.containsKey(id) || collaborators.containsKey(id));
+
+        assertSame(router.leadAgents(), router.agentsFor("term_collab"),
+                "a configured collaborator's terminal must route to the LEAD daemon, not the member "
+                        + "one — its pane is opened by a person, exactly like a lead's");
+    }
+
+    /**
+     * Companion to {@link #collaboratorTerminalRoutesToTheLeadDaemon}: a terminal that is neither a
+     * known lead nor a known collaborator must still route to the member daemon. Without this, a
+     * predicate of {@code _ -> true} would also pass the test above.
+     */
+    @Test
+    void terminalInNeitherMapStillRoutesToTheMemberDaemon() {
+        FakeHerdr lead = new FakeHerdr();
+        FakeHerdr member = new FakeHerdr();
+        Map<String, String> leads = Map.of("term_lead", "primary");
+        Map<String, String> collaborators = Map.of("term_collab", "reviewer-alex");
+        HerdrRouter router = new HerdrRouter(lead, member,
+                id -> leads.containsKey(id) || collaborators.containsKey(id));
+
+        assertSame(router.memberAgents(), router.agentsFor("term_worker"),
+                "a terminal absent from both maps must stay on the member daemon — the fix widens "
+                        + "the predicate, it does not make it unconditionally true");
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/herdr/LeadTabScannerTest.java

@@ -158,9 +158,25 @@ class LeadTabScannerTest {
         return Map.of("lead: opus-5.0", "opus-5.0", "lead: gpt-sol-5.6", "gpt-sol-5.6");
     }
 
+    /** {@code twoLeads()}'s own space — every lead-label fixture below lives here unless noted. */
+    private static final String MAIN_SPACE = "main";
+
+    /** Wraps a flat label → name map under one space, the shape {@link LeadTabScanner} now takes. */
+    private static Map<String, Map<String, String>> inSpace(String space, Map<String, String> labelToName) {
+        return Map.of(space, labelToName);
+    }
+
     private LeadTabScanner scanner(TopologyHerdr herdr, Map<String, String> tabToName,
                                    AtomicLong clock) {
-        return new LeadTabScanner(herdr, tabToName, Set.of("fleetd-workers"), TTL, clock::get);
+        return new LeadTabScanner(herdr, inSpace(MAIN_SPACE, tabToName), Set.of("fleetd-workers"),
+                TTL, clock::get);
+    }
+
+    private LeadTabScanner scannerWithCollaborators(TopologyHerdr herdr, Map<String, String> tabToName,
+                                                     Map<String, String> collaboratorTabToName,
+                                                     AtomicLong clock) {
+        return new LeadTabScanner(herdr, inSpace(MAIN_SPACE, tabToName), collaboratorTabToName,
+                Set.of("fleetd-workers"), TTL, clock::get);
     }
 
     @Test
@@ -198,9 +214,12 @@ class LeadTabScannerTest {
     }
 
     /**
-     * The guard that matters: fleetd labels its own worker tabs, so if a worker space were scanned
-     * a naming accident would promote the fleet. The exclusion is by workspace, not by hoping the
-     * worker template never collides.
+     * Covers the {@code excludedWorkspaceLabels} parameter: a tab in an excluded workspace is never
+     * matched, whatever its label. Production always constructs this class with an empty set (CB-558,
+     * {@code FleetdAssembly}), so this parameter plays no part in the live guard against a worker
+     * tab being mistaken for a lead — that guard is {@code CallerResolver} asking the live
+     * spawned-member roster before any tab map. This test exists because the parameter still exists
+     * and is worth covering on its own terms.
      */
     @Test
     void aTabInAWorkerSpaceIsNeverALeadEvenWhenItsLabelMatches() {
@@ -212,6 +231,76 @@ class LeadTabScannerTest {
         assertFalse(scanner(herdr, tabToName, new AtomicLong()).get().containsKey("term_impostor"));
     }
 
+    /**
+     * The shipped shape (CB-558, {@code FleetdAssembly}): production always constructs this class
+     * with an empty {@code excludedWorkspaceLabels}, and a lead's {@code workspace:} default is the
+     * same shared {@code "fleet"} space the members use. A lead tab is still found when it sits in
+     * the exact same workspace as a member-labelled tab — the scanner tells them apart by the exact
+     * tab label, not by which workspace either one is in.
+     */
+    @Test
+    void aLeadIsDiscoveredWhenItsWorkspaceIsTheSameAsTheMemberWorkspace() {
+        TopologyHerdr herdr = new TopologyHerdr()
+                .workspace("w1", "fleet")
+                .tab("w1:t1", "w1", "lead: opus-5.0")
+                .tab("w1:t2", "w1", "worker: gx10 #1")
+                .pane("w1:p1", "w1:t1", "term_opus")
+                .pane("w1:p2", "w1:t2", "term_worker");
+        LeadTabScanner s = new LeadTabScanner(herdr,
+                inSpace("fleet", Map.of("lead: opus-5.0", "opus-5.0")), Set.of(), TTL,
+                new AtomicLong()::get);
+
+        assertEquals("opus-5.0", s.get().get("term_opus"),
+                "a lead sharing the members' workspace is still discovered — the label, not the "
+                        + "workspace, is what matches it");
+    }
+
+    // ── fleetd #770: space is the uniqueness boundary, not the label alone ──────────────────────
+
+    /**
+     * Two leads can share the exact same label (the fixed {@code lead} tab label) as long as they
+     * sit in different spaces — each tab resolves to its own space's lead, never the other one's.
+     */
+    @Test
+    void aTabLabelledLeadResolvesToItsOwnSpacesLeadNotTheOtherSpaces() {
+        TopologyHerdr herdr = new TopologyHerdr()
+                .workspace("wa", "space-a")
+                .workspace("wb", "space-b")
+                .tab("wa:t1", "wa", "lead")
+                .tab("wb:t1", "wb", "lead")
+                .pane("wa:p1", "wa:t1", "term_a")
+                .pane("wb:p1", "wb:t1", "term_b");
+        Map<String, Map<String, String>> leadLabelsBySpace = Map.of(
+                "space-a", Map.of("lead", "alpha"),
+                "space-b", Map.of("lead", "beta"));
+        LeadTabScanner s = new LeadTabScanner(herdr, leadLabelsBySpace, Set.of(), TTL,
+                new AtomicLong()::get);
+
+        Map<String, String> leads = s.get();
+        assertEquals("alpha", leads.get("term_a"), "space-a's tab must resolve to space-a's lead");
+        assertEquals("beta", leads.get("term_b"), "space-b's tab must resolve to space-b's lead");
+    }
+
+    /**
+     * A lead's deprecated legacy {@code tab:} label is still matched, but only within that lead's
+     * own configured space — exactly the shape {@code FleetdAssembly} builds via {@code
+     * Leader.acceptedLabels()}.
+     */
+    @Test
+    void aLegacyTabLabelStillResolvesWithinItsOwnSpace() {
+        TopologyHerdr herdr = new TopologyHerdr()
+                .workspace("w1", "fleet")
+                .tab("w1:t1", "w1", "lead: opus")
+                .pane("w1:p1", "w1:t1", "term_opus");
+        Map<String, Map<String, String>> leadLabelsBySpace =
+                Map.of("fleet", Map.of("lead", "opus", "lead: opus", "opus"));
+        LeadTabScanner s = new LeadTabScanner(herdr, leadLabelsBySpace, Set.of(), TTL,
+                new AtomicLong()::get);
+
+        assertEquals("opus", s.get().get("term_opus"),
+                "the deprecated tab label must still resolve this lead within its own space");
+    }
+
     @Test
     void aLabelWithNoConfiguredEntryIsIgnored() {
         TopologyHerdr herdr = new TopologyHerdr().workspace("w1", "main")
@@ -464,4 +553,75 @@ class LeadTabScannerTest {
 
         assertEquals(afterFirst, herdr.calls, "the failure path must be rate-limited too");
     }
+
+    // ── fleetd #669 Unit D: a collaborator tab is matched the same way as a lead tab, one pass ─────
+
+    @Test
+    void aConfiguredCollaboratorTabIsReportedByCollaboratorsNotByGet() {
+        TopologyHerdr herdr = new TopologyHerdr()
+                .workspace("w1", "main")
+                .tab("w1:t1", "w1", "collab: ops")
+                .pane("w1:p1", "w1:t1", "term_ops");
+        LeadTabScanner s = scannerWithCollaborators(herdr, Map.of(), Map.of("collab: ops", "ops"),
+                new AtomicLong());
+
+        assertEquals(Map.of("term_ops", "ops"), s.collaborators(),
+                "a collaborator tab is matched exactly like a lead tab");
+        assertEquals(Map.of(), s.get(), "a collaborator tab must never also appear as a lead");
+    }
+
+    /**
+     * Criterion 4, scanner level: a collaborator tab that is labelled but runs no agent is not
+     * reported — the same #359 liveness cross-check a lead tab gets.
+     */
+    @Test
+    void aDeadCollaboratorTabIsNotReported() {
+        TopologyHerdr herdr = new TopologyHerdr()
+                .workspace("w1", "main")
+                .tab("w1:t1", "w1", "collab: ops")
+                .pane("w1:p1", "w1:t1", "term_ops")
+                .deadAgent("w1:t1");
+        LeadTabScanner s = scannerWithCollaborators(herdr, Map.of(), Map.of("collab: ops", "ops"),
+                new AtomicLong());
+
+        assertFalse(s.collaborators().containsKey("term_ops"),
+                "a dead collaborator tab must never resolve as a live collaborator");
+    }
+
+    /**
+     * Both kinds are matched in a single pass over the same tab list — not a second scanner, not a
+     * second scan. Proven by herdr call count: scanning one lead tab and one collaborator tab in the
+     * same instance costs exactly as many calls as scanning two lead tabs in {@link #twoLeads()}.
+     */
+    @Test
+    void leadsAndCollaboratorsAreMatchedInOnePassOverTheSameScan() {
+        TopologyHerdr oneOfEach = new TopologyHerdr()
+                .workspace("w1", "main")
+                .tab("w1:t1", "w1", "lead: opus-5.0")
+                .tab("w1:t2", "w1", "collab: ops")
+                .pane("w1:p1", "w1:t1", "term_opus")
+                .pane("w1:p2", "w1:t2", "term_ops");
+        LeadTabScanner s = scannerWithCollaborators(oneOfEach, Map.of("lead: opus-5.0", "opus-5.0"),
+                Map.of("collab: ops", "ops"), new AtomicLong());
+
+        assertEquals(Map.of("term_opus", "opus-5.0"), s.get());
+        assertEquals(Map.of("term_ops", "ops"), s.collaborators());
+
+        TopologyHerdr twoLeadsBaseline = twoLeads();
+        scanner(twoLeadsBaseline, twoLeadsConfigured(), new AtomicLong()).get();
+
+        assertEquals(twoLeadsBaseline.calls, oneOfEach.calls,
+                "one lead tab + one collaborator tab must cost exactly as many herdr calls as two "
+                        + "lead tabs — proof this is one pass, not a second scan");
+    }
+
+    /** Regression: with no collaborators configured, every existing lead-only behaviour is unchanged. */
+    @Test
+    void anEmptyCollaboratorMapLeavesCollaboratorsEmptyAndGetUnaffected() {
+        LeadTabScanner s = scannerWithCollaborators(twoLeads(), twoLeadsConfigured(), Map.of(),
+                new AtomicLong());
+
+        assertEquals(Map.of(), s.collaborators());
+        assertEquals(Map.of("term_opus", "opus-5.0", "term_gpt", "gpt-sol-5.6"), s.get());
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/herdr/PromptBoxTest.java

@@ -0,0 +1,132 @@
+package dev.ltms.fleet.herdr;
+
+import org.junit.jupiter.api.Test;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/** Reading a Claude Code input box, so a paste-and-submit delivery never submits the operator's draft. */
+class PromptBoxTest {
+
+    private static final String EMPTY = FakeHerdr.IDLE_PROMPT_CARET;
+    private static final String DRAFTED = FakeHerdr.DRAFTED_PROMPT_CARET;
+
+    // --- pure classification -------------------------------------------------
+
+    @Test
+    void anEmptyCaretLineIsAnEmptyBox() {
+        assertEquals(new PromptBox.Reading(PromptBox.State.EMPTY, 0), PromptBox.classify(EMPTY));
+    }
+
+    @Test
+    void aCaretLineHoldingTextIsADraftAndCountsItsCharacters() {
+        PromptBox.Reading reading = PromptBox.classify(DRAFTED);
+        assertEquals(PromptBox.State.DRAFT, reading.state());
+        assertEquals("yes,sendittolead:opus".length(), reading.characters(),
+                "padding does not count — only what the operator typed");
+    }
+
+    @Test
+    void aBorderedBoxIsReadToo() {
+        assertEquals(PromptBox.State.EMPTY, PromptBox.classify(FakeHerdr.IDLE_PROMPT_BOX).state(),
+                "an older TUI draws a bordered box, and its panes must still be readable");
+        assertEquals(PromptBox.State.DRAFT, PromptBox.classify(FakeHerdr.DRAFTED_PROMPT_BOX).state());
+    }
+
+    @Test
+    void aSingleTypedCharacterIsADraft() {
+        assertEquals(PromptBox.State.DRAFT, PromptBox.classify("❯ f").state());
+        assertEquals(PromptBox.State.DRAFT, PromptBox.classify("│ > f                   │").state());
+    }
+
+    @Test
+    void aCursorBlockInAnOtherwiseEmptyBoxIsEmpty() {
+        assertEquals(PromptBox.State.EMPTY, PromptBox.classify("❯ █").state(),
+                "a terminal capture may leave the cursor cell in an empty box");
+        assertEquals(PromptBox.State.EMPTY, PromptBox.classify("│ > █                   │").state());
+    }
+
+    @Test
+    void theBoxLineIsReadToItsEndWhateverFollowsIt() {
+        assertEquals(PromptBox.State.DRAFT, PromptBox.classify("❯ half a line\n  ⏵⏵ auto mode on").state());
+        assertEquals(PromptBox.State.EMPTY, PromptBox.classify("❯\n  ⏵⏵ auto mode on").state());
+    }
+
+    @Test
+    void theLastBoxLineOnThePaneIsTheLiveOne() {
+        assertEquals(PromptBox.State.DRAFT,
+                PromptBox.classify("❯ an earlier prompt\n⏺ its answer\n❯ typing now").state(),
+                "the detection region carries scrollback, so earlier prompts sit above the live box");
+        assertEquals(PromptBox.State.EMPTY,
+                PromptBox.classify("❯ an earlier prompt\n⏺ its answer\n❯").state());
+    }
+
+    @Test
+    void aMarkerPartWayAlongALineIsNotABox() {
+        assertEquals(PromptBox.State.UNREADABLE, PromptBox.classify("⏺ type ❯ to get a prompt").state(),
+                "a caret the operator quoted is transcript text, not an input box");
+        assertEquals(PromptBox.State.EMPTY, PromptBox.classify("⏺ type ❯ to get a prompt\n❯").state(),
+                "and it must not shadow the real box further down");
+    }
+
+    @Test
+    void aPaneWithNoBoxIsUnreadable() {
+        assertEquals(PromptBox.State.UNREADABLE, PromptBox.classify("garbled ansi noise").state());
+        assertEquals(PromptBox.State.UNREADABLE, PromptBox.classify("").state());
+        assertEquals(PromptBox.State.UNREADABLE, PromptBox.classify(null).state());
+    }
+
+    @Test
+    void aGeneratingTurnIsUnreadableEvenWithAnEmptyBox() {
+        assertEquals(PromptBox.State.UNREADABLE,
+                PromptBox.classify(EMPTY + "\n  ✳ Thinking… (12s · esc to interrupt)").state());
+    }
+
+    @Test
+    void aGeneratingMarkerInScrollbackAboveTheBoxDoesNotMakeThePaneUnreadable() {
+        assertEquals(PromptBox.State.EMPTY,
+                PromptBox.classify("  ✳ Thinking… (12s · esc to interrupt)\n⏺ done\n" + EMPTY).state(),
+                "that marker survives in scrollback, and holding on it would hold every delivery forever");
+    }
+
+    // --- the gate ------------------------------------------------------------
+
+    @Test
+    void anEmptyBoxClearsTheGateAndReadsTheDetectionRegion() {
+        FakeHerdr herdr = new FakeHerdr().detectionText(EMPTY);
+
+        assertTrue(new PromptBox(new AgentControl(herdr)).clearToSubmit("term_a"));
+
+        @SuppressWarnings("unchecked")
+        var params = (java.util.Map<String, Object>) herdr.lastCall("agent.read").params();
+        assertEquals("detection", params.get("source"),
+                "the input box is drawn in the detection region, not in transcript scrollback");
+    }
+
+    @Test
+    void aDraftedBoxHoldsTheGate() {
+        assertFalse(new PromptBox(new AgentControl(new FakeHerdr().detectionText(DRAFTED))).clearToSubmit("term_a"));
+    }
+
+    @Test
+    void anUnreadablePaneHoldsTheGate() {
+        assertFalse(new PromptBox(new AgentControl(new FakeHerdr().detectionText("garbled"))).clearToSubmit("term_a"));
+    }
+
+    @Test
+    void aFailedReadHoldsTheGate() {
+        assertFalse(new PromptBox(new AgentControl(new FakeHerdr().healthy(false))).clearToSubmit("term_a"),
+                "a pane this cannot read must never be pasted into");
+    }
+
+    @Test
+    void theGateClearsAgainOnceTheBoxEmpties() {
+        FakeHerdr herdr = new FakeHerdr().detectionText(DRAFTED);
+        PromptBox box = new PromptBox(new AgentControl(herdr));
+
+        assertFalse(box.clearToSubmit("term_a"));
+        herdr.detectionText(EMPTY);
+        assertTrue(box.clearToSubmit("term_a"));
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/inject/BackendOutageFlowTest.java

@@ -89,6 +89,11 @@ class BackendOutageFlowTest {
                 return MAPPER.createObjectNode().set("agent", MAPPER.createObjectNode()
                         .put("terminal_id", "term_primary").put("agent_status", "idle"));
             }
+            if ("agent.read".equals(method)) {
+                // The lead-nudge paths read the input box before pasting into it.
+                return MAPPER.createObjectNode().set("read",
+                        MAPPER.createObjectNode().put("text", FakeHerdr.IDLE_PROMPT_CARET));
+            }
             if ("agent.prompt".equals(method)) {
                 prompts.add(params);
                 sendLatch.countDown();

fleetd/src/test/java/dev/ltms/fleet/inject/InjectorTest.java

@@ -59,6 +59,17 @@ class InjectorTest {
         assertEquals(List.of("hello"), sent());
     }
 
+    @Test
+    void deliveringToAMemberReadsNoPane() {
+        // No human types into a spawned member's pane, so its delivery path must not pay for a
+        // prompt-box read the way a lead's nudge paths do.
+        injector.enqueue(T, "task", TestTurnTokens.inert(T));
+        injector.onStatus(T, AgentStatus.IDLE);
+
+        assertEquals(List.of("task"), sent());
+        assertFalse(herdr.called("agent.read"), "a member delivery must not read its pane");
+    }
+
     @Test
     void holdsDeliveryUntilTheWorkerIsAvailable() {
         // CB-113: idle alone is not enough — hold until the worker's MCP is connected (ready).

fleetd/src/test/java/dev/ltms/fleet/lead/LeadLauncherTest.java

@@ -1,10 +1,16 @@
 package dev.ltms.fleet.lead;
 
+import ch.qos.logback.classic.Level;
+import ch.qos.logback.classic.Logger;
+import ch.qos.logback.classic.spi.ILoggingEvent;
+import ch.qos.logback.core.read.ListAppender;
 import dev.ltms.fleet.config.FleetConfig;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.FakeHerdr;
+import dev.ltms.fleet.herdr.ResilientAgentLaunch;
 import dev.ltms.fleet.herdr.WorkspaceControl;
 import org.junit.jupiter.api.Test;
+import org.slf4j.LoggerFactory;
 
 import java.util.LinkedHashMap;
 import java.util.List;
@@ -64,6 +70,21 @@ class LeadLauncherTest {
         return new LeadLauncher(new AgentControl(herdr), new WorkspaceControl(herdr), cfg);
     }
 
+    /** As {@link #launcher}, plus a fast no-op sleeper so a busy-retry test never real-sleeps. */
+    private static LeadLauncher fastLauncher(FakeHerdr herdr, FleetConfig cfg) {
+        return new LeadLauncher(new AgentControl(herdr), new WorkspaceControl(herdr), cfg, () -> { });
+    }
+
+    /** {@link #opusProfile()} with one argv element long enough to overflow the pane line limit. */
+    private static FleetConfig.Profile hugeArgvProfile() {
+        return new FleetConfig.Profile(
+                "opus", null, "claude-opus-5", null, "FLEETD_WORKER_TOKEN",
+                List.of("ccs", "x".repeat(1500)), "tab", "fleet", null,
+                "http://127.0.0.1:8765/mcp", null, null,
+                null, null, null,
+                Map.of("CLAUDE_CODE_AUTO_COMPACT_WINDOW", "300000"), null, null, true, null);
+    }
+
     @SuppressWarnings("unchecked")
     private static List<String> startedArgs(FakeHerdr herdr) {
         return (List<String>) ((Map<String, Object>) herdr.lastCall("agent.start").params()).get("args");
@@ -86,16 +107,20 @@ class LeadLauncherTest {
 
         assertEquals(1, launcher(herdr, configWith(lead("opus", "lead: opus", 1))).ensureLeads());
         assertTrue(herdr.called("agent.start"), "a lead must actually be started");
-        assertEquals("lead-opus", startedName(herdr));
+        // fleetd #727: the name carries a per-process nonce and a per-start sequence number — the
+        // same unique-naming scheme HerdrPeerLauncher uses for members — rather than the fixed
+        // "lead-opus" a stale registry entry could block a legitimate relaunch under.
+        assertTrue(startedName(herdr).matches("lead-opus-[0-9a-f]{6}-\\d+"),
+                "name is lead-<name>-<nonce>-<seq>: " + startedName(herdr));
     }
 
-    /** The tab is labelled with the configured `tab:` so the scanner finds the lead on the next resolve. */
+    /** The tab is labelled with the fixed lead tab label so the scanner finds the lead on the next resolve. */
     @Test
-    void labelsTheTabWithTheConfiguredTabValue() {
+    void labelsTheTabWithTheFixedLeadTabLabel() {
         FakeHerdr herdr = new FakeHerdr();
         launcher(herdr, configWith(lead("opus", "lead: opus", 1))).ensureLeads();
 
-        assertEquals("lead: opus",
+        assertEquals("lead",
                 ((Map<?, ?>) herdr.lastCall("tab.rename").params()).get("label"));
     }
 
@@ -123,6 +148,25 @@ class LeadLauncherTest {
         assertFalse(herdr.called("tab.close"), "a labelled tab WITH a live agent must never be closed");
     }
 
+    /**
+     * The tab a live lead actually sits in still carries its deprecated legacy {@code tab:} label,
+     * not the fixed {@code lead} tab label a freshly auto-launched instance would get. Counting must
+     * still recognise it as the live lead via {@link FleetConfig.Leader#acceptedLabels()}, or a
+     * daemon restart would read it as missing and launch a second orchestrator next to the first.
+     */
+    @Test
+    void aLiveLeadInALegacyLabelledTabIsCountedSoNothingIsLaunched() {
+        FakeHerdr herdr = new FakeHerdr()
+                .withWorkspace("wL", "fleet")
+                .withTab("wL", "wL:t1", "lead: opus")
+                .withAgent("lead-opus", "term_lead", "wL:p1", "wL:t1");
+
+        assertEquals(0, launcher(herdr, configWith(lead("opus", "lead: opus", 1))).ensureLeads(),
+                "the legacy-labelled live lead must be counted — nothing may be launched");
+        assertFalse(herdr.called("agent.start"),
+                "a tab label fixed to a constant must not blind the count to a legacy-labelled lead");
+    }
+
     /**
      * The reason liveness is not "does the label exist". A tab left labelled by a session that has
      * since died must not block the relaunch, or one crash disables auto-launch permanently.
@@ -238,7 +282,7 @@ class LeadLauncherTest {
         assertFalse(herdr.called("tab.close"), "a tab running an agent again must never be closed");
         assertFalse(herdr.called("agent.start"), "the lead is live again — nothing to relaunch");
         assertEquals("wL:t1", ((Map<?, ?>) herdr.lastCall("tab.rename").params()).get("tab_id"));
-        assertEquals("lead: opus", ((Map<?, ?>) herdr.lastCall("tab.rename").params()).get("label"),
+        assertEquals("lead", ((Map<?, ?>) herdr.lastCall("tab.rename").params()).get("label"),
                 "the pending-close flag must be cleared once the tab is confirmed live again");
     }
 
@@ -260,7 +304,7 @@ class LeadLauncherTest {
     @Test
     void aHandOpenedLeadWithTheConfiguredTabLabelCountsAsLive() {
         FakeHerdr herdr = new FakeHerdr()
-                .withWorkspace("wX", "main")
+                .withWorkspace("wX", "fleet")
                 .withTab("wX", "wX:t1", "lead: opus")
                 .withAgent("hand-opened", "term_hand", "wX:p1", "wX:t1");
 
@@ -436,4 +480,247 @@ class LeadLauncherTest {
         assertEquals(0, launcher(herdr, cfg).ensureLeads());
         assertTrue(herdr.calls.isEmpty(), "nothing declared ⇒ nothing scanned");
     }
+
+    // ── fleetd #727: the same three launch protections every member spawn gets ───────────────────
+
+    /**
+     * A freshly created pane may not have redrawn its prompt yet, so herdr answers
+     * {@code agent_pane_busy}. The lead launch must wait it out rather than fail on the first miss
+     * — exactly the retry {@code HerdrPeerLauncher} already gives every member.
+     */
+    @Test
+    void aLeadLaunchRetriesWhileTheSeedShellBoots() {
+        FakeHerdr herdr = new FakeHerdr().agentPaneBusyTimes(2);
+
+        assertEquals(1, fastLauncher(herdr, configWith(lead("opus", "lead: opus", 1))).ensureLeads(),
+                "the lead must still start once the shell is ready");
+        assertEquals(3, herdr.calls.stream().filter(c -> c.method().equals("agent.start")).count(),
+                "two busy rejections, then the successful start");
+    }
+
+    /**
+     * The busy retry is bounded, not an infinite poll. If the pane never becomes ready the launch
+     * must eventually give up and log the failure, not hang the daemon's reconcile loop forever —
+     * proven here by a budget that would still be busy on attempt
+     * {@value ResilientAgentLaunch#SHELL_READY_RETRIES} and a call count that stops exactly there.
+     */
+    @Test
+    void aLeadLaunchGivesUpAfterTheBoundedBusyBudgetRatherThanLoopingForever() {
+        FakeHerdr herdr = new FakeHerdr().agentPaneBusyTimes(999);
+
+        assertEquals(0, fastLauncher(herdr, configWith(lead("opus", "lead: opus", 1))).ensureLeads(),
+                "a pane that never becomes ready must not be reported as a started lead");
+        assertEquals(ResilientAgentLaunch.SHELL_READY_RETRIES,
+                herdr.calls.stream().filter(c -> c.method().equals("agent.start")).count(),
+                "the retry budget is bounded: it stops after exactly SHELL_READY_RETRIES attempts");
+    }
+
+    /**
+     * herdr refuses a duplicate agent {@code name} outright. A stale registry entry — a crashed
+     * lead session, or a name the registry has not yet released — must not permanently block a
+     * legitimate relaunch, so each retry attempt carries a fresh per-start name.
+     */
+    @Test
+    @SuppressWarnings("unchecked")
+    void aLeadLaunchRetriesUnderAFreshNameWhenTheOldNameIsStillTaken() {
+        FakeHerdr herdr = new FakeHerdr().agentNameTakenTimes(2);
+
+        assertEquals(1, launcher(herdr, configWith(lead("opus", "lead: opus", 1))).ensureLeads(),
+                "the lead must still start once a free name is found");
+        List<String> names = herdr.calls.stream()
+                .filter(c -> c.method().equals("agent.start"))
+                .map(c -> ((Map<String, Object>) c.params()).get("name").toString())
+                .toList();
+        assertEquals(3, names.size(), "2 rejected + 1 success");
+        assertEquals(3, Set.copyOf(names).size(), "each attempt must use a distinct name");
+    }
+
+    /**
+     * The name-collision retry is bounded too. If the name is taken on every attempt, the launch
+     * must give up rather than keep minting new names forever — the per-start naming scheme means
+     * every failed attempt was refused outright by herdr (no process exists under a name herdr
+     * refused), so a bounded, exhausted retry never leaves a second process running: nothing is
+     * running at all.
+     */
+    @Test
+    void aLeadLaunchNeverEndsUpWithASecondProcessWhenTheNameStaysTaken() {
+        FakeHerdr herdr = new FakeHerdr().agentNameTakenTimes(999);
+
+        assertEquals(0, launcher(herdr, configWith(lead("opus", "lead: opus", 1))).ensureLeads(),
+                "a name that is never free must not be reported as a started lead");
+        assertEquals(ResilientAgentLaunch.NAME_RETRIES,
+                herdr.calls.stream().filter(c -> c.method().equals("agent.start")).count(),
+                "the retry budget is bounded: it stops after exactly NAME_RETRIES attempts, "
+                        + "never racing a duplicate into existence");
+    }
+
+    /**
+     * fleetd #220/#727: herdr types the launch command into the pane as one line, and a pty line
+     * buffer holds only 1024 bytes — past that the tail is dropped with no error at all, and the
+     * backend exits on a mangled argument. The lead launch must refuse an over-long command outright
+     * rather than let it be typed and silently truncated.
+     */
+    @Test
+    void anOverlongLeadArgvIsRefusedRatherThanTypedAndTruncated() {
+        FakeHerdr herdr = new FakeHerdr();
+        Logger logger = (Logger) LoggerFactory.getLogger(LeadLauncher.class);
+        ListAppender<ILoggingEvent> appender = new ListAppender<>();
+        appender.start();
+        logger.addAppender(appender);
+
+        int started;
+        try {
+            started = launcher(herdr, configWith(lead("opus", "lead: opus", 1), hugeArgvProfile()))
+                    .ensureLeads();
+        } finally {
+            logger.detachAppender(appender);
+        }
+
+        assertEquals(0, started, "an over-long command must never be reported as a started lead");
+        assertFalse(herdr.called("agent.start"),
+                "nothing may be started — a truncated command is worse than no spawn");
+        String warn = appender.list.stream()
+                .filter(e -> e.getLevel().equals(Level.WARN))
+                .map(ILoggingEvent::getFormattedMessage)
+                .filter(m -> m.contains("failed to launch"))
+                .findFirst()
+                .orElseThrow(() -> new AssertionError("expected a WARN naming the launch failure: "
+                        + appender.list));
+        assertTrue(warn.contains("1024"), "names the limit: " + warn);
+        assertTrue(warn.contains("opus"), "names the profile: " + warn);
+        assertTrue(warn.contains("x".repeat(60)), "names the culprit argument: " + warn);
+    }
+
+    // ── fleetd #726 unit 1: the single-lead relaunch seam ─────────────────────────────────────
+
+    /**
+     * The returned agent's {@code terminalId()}/{@code paneId()} are the ones the fake
+     * {@code AgentControl} actually started — not a coincidental field left over from the caller.
+     * {@code paneId()} echoes the exact {@code pane_id} the launch's own {@code agent.start} call
+     * carried (protocol 19: the agent starts into the pane it is asked to), and {@code
+     * terminalId()} is herdr's own generated id, which the fake always shapes as {@code
+     * term_new_<n>}.
+     */
+    @Test
+    void relaunchReturnsTheStartedAgent() {
+        FakeHerdr herdr = new FakeHerdr();
+
+        dev.ltms.fleet.herdr.Agent started =
+                launcher(herdr, configWith(lead("opus", "lead: opus", 1))).relaunch("opus");
+
+        assertNotNull(started, "a launchable, configured lead must start");
+        Object startedPaneIdParam = ((Map<?, ?>) herdr.lastCall("agent.start").params()).get("pane_id");
+        assertEquals(startedPaneIdParam, started.paneId(),
+                "paneId() must be the pane the agent.start call actually targeted");
+        assertTrue(started.terminalId() != null && started.terminalId().startsWith("term_new_"),
+                "terminalId() must be herdr's own generated id: " + started.terminalId());
+    }
+
+    /** The new tab is labelled with the fixed lead tab label, and AFTER the start. */
+    @Test
+    void relaunchLabelsTheNewTabAfterStarting() {
+        FakeHerdr herdr = new FakeHerdr();
+
+        dev.ltms.fleet.herdr.Agent started =
+                launcher(herdr, configWith(lead("opus", "lead: opus", 1))).relaunch("opus");
+
+        assertNotNull(started);
+        assertEquals("lead", ((Map<?, ?>) herdr.lastCall("tab.rename").params()).get("label"));
+
+        int startIndex = indexOfLastCall(herdr, "agent.start");
+        int renameIndex = indexOfLastCall(herdr, "tab.rename");
+        assertTrue(renameIndex > startIndex,
+                "the tab must be renamed AFTER the start succeeds, not before: start=" + startIndex
+                        + " rename=" + renameIndex);
+    }
+
+    private static int indexOfLastCall(FakeHerdr herdr, String method) {
+        int idx = -1;
+        List<FakeHerdr.Call> calls = herdr.calls;
+        for (int i = 0; i < calls.size(); i++) {
+            if (calls.get(i).method().equals(method)) {
+                idx = i;
+            }
+        }
+        return idx;
+    }
+
+    @Test
+    void relaunchOfAnUnknownLeadNameReturnsNullAndStartsNothing() {
+        FakeHerdr herdr = new FakeHerdr();
+
+        dev.ltms.fleet.herdr.Agent started =
+                launcher(herdr, configWith(lead("opus", "lead: opus", 1))).relaunch("not-declared");
+
+        assertNull(started);
+        assertFalse(herdr.called("agent.start"));
+        assertFalse(herdr.called("workspace.create"));
+        assertFalse(herdr.called("tab.create"));
+    }
+
+    @Test
+    void relaunchOfARecogniseOnlyLeadReturnsNullAndStartsNothing() {
+        FakeHerdr herdr = new FakeHerdr();
+
+        dev.ltms.fleet.herdr.Agent started =
+                launcher(herdr, configWith(lead(null, "lead: dead", 1))).relaunch("opus");
+
+        assertNull(started);
+        assertFalse(herdr.called("agent.start"));
+    }
+
+    @Test
+    void relaunchWithAnUnconfiguredProfileReturnsNullAndStartsNothing() {
+        FakeHerdr herdr = new FakeHerdr();
+
+        dev.ltms.fleet.herdr.Agent started =
+                launcher(herdr, configWith(lead("nope", "lead: opus", 1))).relaunch("opus");
+
+        assertNull(started);
+        assertFalse(herdr.called("agent.start"));
+    }
+
+    /**
+     * The outer retry {@link LeadLauncher#relaunch(String)} owns, separate from {@code
+     * ResilientAgentLaunch}'s internal {@code agent_name_taken} retry: a failed attempt must not
+     * be the end of the whole relaunch. Each of the first two attempts exhausts {@code
+     * ResilientAgentLaunch.NAME_RETRIES} name attempts (every one of them rejected), so each
+     * attempt's own tab is created and then closed; the third attempt's first name is free.
+     */
+    @Test
+    void relaunchRetriesTheWholeAttemptAndSucceedsOnTheThird() {
+        FakeHerdr herdr = new FakeHerdr()
+                .agentNameTakenTimes(2 * ResilientAgentLaunch.NAME_RETRIES);
+
+        dev.ltms.fleet.herdr.Agent started =
+                fastLauncher(herdr, configWith(lead("opus", "lead: opus", 1))).relaunch("opus");
+
+        assertNotNull(started, "the third attempt's first name is free — it must succeed");
+        assertEquals(3, herdr.calls.stream().filter(c -> c.method().equals("tab.create")).count(),
+                "one tab per attempt: three attempts");
+        assertEquals(2, herdr.calls.stream().filter(c -> c.method().equals("tab.close")).count(),
+                "the two failed attempts' tabs must be closed");
+    }
+
+    /**
+     * Every attempt fails outright (a herdr error {@code ResilientAgentLaunch} does not retry at
+     * all) — {@link LeadLauncher#relaunch(String)} must give up after exactly {@code
+     * RELAUNCH_ATTEMPTS} and must not leak any of the tabs it created along the way.
+     */
+    @Test
+    void relaunchGivesUpAfterExactlyRelaunchAttemptsAndLeaksNoTab() {
+        FakeHerdr herdr = new FakeHerdr().agentStartFailsWith("some_other_error");
+
+        dev.ltms.fleet.herdr.Agent started =
+                fastLauncher(herdr, configWith(lead("opus", "lead: opus", 1))).relaunch("opus");
+
+        assertNull(started, "every attempt failed — relaunch must give up, not hang or guess");
+        assertEquals(LeadLauncher.RELAUNCH_ATTEMPTS,
+                herdr.calls.stream().filter(c -> c.method().equals("agent.start")).count(),
+                "exactly RELAUNCH_ATTEMPTS attempts, no more, no fewer");
+        long tabsCreated = herdr.calls.stream().filter(c -> c.method().equals("tab.create")).count();
+        long tabsClosed = herdr.calls.stream().filter(c -> c.method().equals("tab.close")).count();
+        assertEquals(LeadLauncher.RELAUNCH_ATTEMPTS, tabsCreated);
+        assertEquals(tabsCreated, tabsClosed, "every tab this method created must be closed — no leaks");
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/mcp/FleetMcpAuthzTest.java

@@ -17,6 +17,7 @@ import dev.ltms.fleet.metrics.Metrics;
 import dev.ltms.fleet.msg.InMemoryReplyInbox;
 import dev.ltms.fleet.msg.MessageService;
 import dev.ltms.fleet.msg.Rendezvous;
+import dev.ltms.fleet.peer.MemberRole;
 import dev.ltms.fleet.session.FakeWorktrees;
 import dev.ltms.fleet.session.SessionManager;
 import dev.ltms.fleet.member.ClaudeCodeLauncher;
@@ -63,6 +64,17 @@ class FleetMcpAuthzTest {
 
     /** A fully wired FleetMcp on fakes — constructing it is itself part of what is under test. */
     private FleetMcp mcp(boolean enforce) {
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> 999_999);
+        return mcp(enforce, CallerResolver.withLeadsAndMembers(identity, false, null,
+                Map::of, new MemberRegistry(null)));
+    }
+
+    /**
+     * As {@link #mcp(boolean)}, with an explicit {@link CallerResolver} — so a test can wire known
+     * leads/collaborators and drive {@code denyFor}'s real {@code knownLeadOrCollaborator()}
+     * classifier instead of the default empty one.
+     */
+    private FleetMcp mcp(boolean enforce, CallerResolver callers) {
         FleetConfig.Profile cfg = new FleetConfig.Profile(
                 "ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
                 "tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
@@ -83,9 +95,7 @@ class FleetMcpAuthzTest {
         // to be omitted to reach "legacy" is now always real, and AuthorizationMode is the
         // separate, explicit choice that governs enforcement.
         mcp = new FleetMcp(messages, workers, sessions, identity, sessions.asPresence(),
-                new PrimaryRegistry(null),
-                CallerResolver.withLeadsAndMembers(identity, false, null,
-                Map::of, new MemberRegistry(null)),
+                new PrimaryRegistry(null), callers,
                 enforce ? FleetMcp.AuthorizationMode.ENFORCED : FleetMcp.AuthorizationMode.UNENFORCED,
                 metrics, FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), null, FleetMcp.OutageSource.none(),
@@ -97,6 +107,7 @@ class FleetMcpAuthzTest {
     private static final Principal WORKER_A = Principal.worker("term_a", 200);
     private static final Principal ANON = Principal.anonymous();
     private static final Principal ARCH_DESIGN = Principal.architect("lead-designer", "term_design", 400);
+    private static final Principal COLLABORATOR = Principal.collaborator("ops", "term_collab", 600);
 
     // --- the table, enforced on THIS path too ---------------------------------------------------
 
@@ -156,6 +167,45 @@ class FleetMcpAuthzTest {
         }
     }
 
+    /**
+     * fleetd #669 Unit A: {@code SEND} is split into three actions ({@link Authz.Action#SEND},
+     * {@link Authz.Action#COORD_SEND}, {@link Authz.Action#ANSWER}), each carrying the same grant
+     * the one undivided action gave. An architect holds all three, exactly as it held the one.
+     */
+    @Test
+    void anArchitectMayUseAllThreeSendShapesOverMcp() {
+        FleetMcp m = mcp(true);
+        for (Authz.Action a : new Authz.Action[]{Authz.Action.SEND, Authz.Action.COORD_SEND,
+                Authz.Action.ANSWER}) {
+            assertNull(m.denyFor(ARCH_DESIGN, a, "term_a"),
+                    a + " carries the same grant the undivided SEND action gave an architect");
+        }
+    }
+
+    /** The other half of the same split: a worker is excluded from all three, as it was from one. */
+    @Test
+    void aWorkerMayNotUseAnySendShapeOverMcp() {
+        FleetMcp m = mcp(true);
+        for (Authz.Action a : new Authz.Action[]{Authz.Action.SEND, Authz.Action.COORD_SEND,
+                Authz.Action.ANSWER}) {
+            McpSchema.CallToolResult denied = m.denyFor(WORKER_A, a, "term_a");
+            assertNotNull(denied, a + " must stay refused to a worker");
+            assertTrue(denied.isError(), "a refusal is returned as an MCP tool error");
+        }
+    }
+
+    /**
+     * fleetd #669 Unit A / #678: {@code TASK_READ} (ticket polling, session status) is split out of
+     * the roster-only {@code READ}, carrying forward the grant the undivided action gave. A worker
+     * still has both — it never gained or lost anything by the split.
+     */
+    @Test
+    void aWorkerKeepsBothReadActionsAfterTheSplit() {
+        FleetMcp m = mcp(true);
+        assertNull(m.denyFor(WORKER_A, Authz.Action.READ, null));
+        assertNull(m.denyFor(WORKER_A, Authz.Action.TASK_READ, null));
+    }
+
     @Test
     void anArchitectMayReplyAndAskOnlyAsItsOwnPaneOverMcp() {
         FleetMcp m = mcp(true);
@@ -187,6 +237,93 @@ class FleetMcpAuthzTest {
                 "the caller IS authenticated — it is just not the right role");
     }
 
+    /**
+     * {@code denyFor} passes the real production classifier, not a test-supplied one — no
+     * terminal is recognised as a configured lead or collaborator, so a collaborator's SEND is
+     * refused over MCP.
+     */
+    @Test
+    void aCollaboratorMayNotSendOverMcpWithTheRealProductionClassifier() {
+        FleetMcp m = mcp(true);
+        McpSchema.CallToolResult denied = m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_lead");
+        assertNotNull(denied, "no terminal is recognised as a lead or collaborator yet");
+        assertTrue(denied.isError());
+    }
+
+    /**
+     * fleetd #669 Unit D: wires a real {@link CallerResolver} with a known lead and a known
+     * collaborator tab, and leaves a spawned member's own terminal recognised by neither map — so a
+     * collaborator's SEND reaches both named peers and is refused for the spawned member's terminal,
+     * over MCP's {@code denyFor}.
+     */
+    @Test
+    void aCollaboratorMaySendToAKnownLeadOrCollaboratorButNotToASpawnedMembersTerminal() {
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> 999_999);
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
+                () -> Map.of("term_lead_known", "lead-x"), new MemberRegistry(null),
+                t -> null, () -> Map.of("term_collab_known", "ops2"));
+        FleetMcp m = mcp(true, callers);
+
+        assertNull(m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_lead_known"));
+        assertNull(m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_collab_known"));
+        assertNotNull(m.denyFor(COLLABORATOR, Authz.Action.SEND, "term_a"),
+                "a spawned member's own terminal must stay unreachable, even once the classifier is real");
+    }
+
+    // --- fleetd #743: the observer SEND matrix, over MCP's denyFor -------------------------------
+
+    private static final Principal OBSERVER = Principal.observer("term_observer", 700);
+
+    /**
+     * Wires one real {@link CallerResolver} that recognises a lead, a collaborator, and a live
+     * spawned worker, leaving "term_other_observer" classified as none of them — so the same
+     * wiring both denies an observer's {@code SEND} to every privileged role and grants it to
+     * another unclassified pane, proving the refusals are the rule and not a missing fixture.
+     */
+    @Test
+    void anObserverMaySendOnlyToAnotherObserverNeverToALeadWorkerOrArchitect() {
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> 999_999);
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
+                () -> Map.of("term_lead_known", "lead-x"), new MemberRegistry(null),
+                t -> "term_a".equals(t) ? MemberRole.DEV : null,
+                () -> Map.of("term_collab_known", "ops2"));
+        FleetMcp m = mcp(true, callers);
+
+        assertNotNull(m.denyFor(OBSERVER, Authz.Action.SEND, "term_lead_known"),
+                "an observer must never reach a lead's terminal");
+        assertNotNull(m.denyFor(OBSERVER, Authz.Action.SEND, "term_collab_known"),
+                "an observer must never reach a collaborator's terminal");
+        assertNotNull(m.denyFor(OBSERVER, Authz.Action.SEND, "term_a"),
+                "an observer must never reach a live spawned member's terminal");
+
+        // CONTROL: the same wiring, the same denyFor call, a target recognised as none of the
+        // three privileged roles above -- this is what proves the three refusals above are the
+        // rule working, not a classifier that refuses every target regardless of what it is.
+        assertNull(m.denyFor(OBSERVER, Authz.Action.SEND, "term_other_observer"),
+                "an observer must reach another pane that resolves as an observer itself");
+    }
+
+    /**
+     * As {@link #anObserverMaySendOnlyToAnotherObserverNeverToALeadWorkerOrArchitect}, for a
+     * terminal bound to a configured architect slot but hosting no live spawned-member session --
+     * the case {@link CallerResolver#resolve} itself treats separately from a live worker/architect.
+     */
+    @Test
+    void anObserverMayNotSendToABoundArchitectSlotEither() {
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> 999_999);
+        MemberRegistry members = new MemberRegistry(new FleetConfig.Fleet(Map.of(),
+                Map.of("lead-designer", new FleetConfig.Slot("sonnet")), Map.of(), Map.of(), null));
+        assertTrue(members.bind("architect:lead-designer", "term_bound_architect"));
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null, Map::of,
+                members, t -> null, Map::of);
+        FleetMcp m = mcp(true, callers);
+
+        assertNotNull(m.denyFor(OBSERVER, Authz.Action.SEND, "term_bound_architect"),
+                "a terminal bound to a configured architect slot must stay unreachable to an observer");
+        // CONTROL: the same wiring, a target the bind above never touched.
+        assertNull(m.denyFor(OBSERVER, Authz.Action.SEND, "term_other_observer"));
+    }
+
     @Test
     void theLegacyConstructorLeavesTheGateOpen() {
         // The 22 pre-existing FleetMcpTest cases rely on no authorization being enforced.
@@ -273,10 +410,10 @@ class FleetMcpAuthzTest {
                         + "the anchors have drifted, this test is not testing what it claims to");
 
         Pattern trailingArg = Pattern.compile(
-                "listFleet\\([^;]*?,\\s*(coordinatorVisibleTo\\(principal\\(exchange\\)\\)|true|false)\\s*\\)\\s*;",
+                "listFleet\\([^;]*?,\\s*(coordinatorVisibleTo\\(principal\\(exchange\\)\\)|true|false)\\s*[,)]",
                 Pattern.DOTALL);
         Matcher m = trailingArg.matcher(handlerBlock);
-        assertTrue(m.find(), "could not locate listFleet(...)'s trailing boolean argument in the "
+        assertTrue(m.find(), "could not locate listFleet(...)'s coordinator boolean argument in the "
                 + "listHandler block -- the call shape changed, update this test's anchor: " + handlerBlock);
         String trailing = m.group(1);
         assertEquals("coordinatorVisibleTo(principal(exchange))", trailing,
@@ -284,6 +421,219 @@ class FleetMcpAuthzTest {
                         + "calling, not pass a literal boolean -- found: " + trailing);
     }
 
+    // --- fleetd #703: who may see fleet_list's collaborators array -------------------------------
+
+    /**
+     * fleetd #703: {@link FleetMcp#collaboratorsVisibleTo} is the whole policy decision for
+     * {@code fleet_list}'s {@code collaborators} array. Visible to exactly the roles that may
+     * {@code SEND} to a named peer -- the primary, an architect, and a collaborator itself -- never
+     * a worker, which holds {@code READ} but can never {@code SEND} to a collaborator, and never an
+     * anonymous caller.
+     */
+    @Test
+    void onlyPrimaryArchitectAndCollaboratorMaySeeTheCollaboratorsArray() {
+        assertTrue(FleetMcp.collaboratorsVisibleTo(PRIMARY), "the primary must see the collaborators array");
+        assertTrue(FleetMcp.collaboratorsVisibleTo(ARCH_DESIGN), "an architect must see the collaborators array");
+        assertTrue(FleetMcp.collaboratorsVisibleTo(COLLABORATOR), "a collaborator must see its own peer roster");
+        assertFalse(FleetMcp.collaboratorsVisibleTo(WORKER_A),
+                "a worker holds READ but can never SEND to a collaborator, so it must not see the array");
+        assertFalse(FleetMcp.collaboratorsVisibleTo(ANON), "authenticated as nothing must not see it either");
+    }
+
+    /**
+     * fleetd #703, same reasoning as {@link #theFleetListHandlerActuallyConsultsCoordinatorVisibleTo}:
+     * the predicate above can be perfectly correct while the one production call site never asks it.
+     * This reads {@code FleetMcp.java}'s own source and asserts the {@code fleet_list} handler's
+     * {@code listFleet(...)} call both threads {@code callers.collaborators()} into the payload and
+     * asks {@code collaboratorsVisibleTo(principal(exchange))} for the visibility flag, rather than a
+     * literal boolean or an empty map.
+     */
+    @Test
+    void theFleetListHandlerActuallyConsultsCollaboratorsVisibleTo() throws Exception {
+        String source = Files.readString(MCP_SOURCE);
+
+        int start = source.indexOf("listHandler =");
+        assertTrue(start >= 0, "could not find the fleet_list handler (listHandler) in " + MCP_SOURCE
+                + " -- the scrape has stopped matching, fix the anchor before trusting this test");
+        int end = source.indexOf("stopHandler =", start);
+        assertTrue(end > start, "could not find the handler declared after listHandler to bound the scrape");
+        String handlerBlock = source.substring(start, end);
+
+        // CONTROL: the block we scraped really does contain a call to listFleet(...) -- if this
+        // fails, the anchors above moved and the assertions below would otherwise pass on nothing.
+        assertTrue(handlerBlock.contains("listFleet("),
+                "control failed: the scraped listHandler block contains no listFleet( call at all -- "
+                        + "the anchors have drifted, this test is not testing what it claims to");
+
+        assertTrue(handlerBlock.contains("callers.collaborators()"),
+                "the fleet_list handler must thread callers.collaborators() into listFleet(...), not an "
+                        + "empty or literal map -- block: " + handlerBlock);
+        assertTrue(handlerBlock.contains("collaboratorsVisibleTo(principal(exchange))"),
+                "the fleet_list handler must ask collaboratorsVisibleTo(principal(exchange)) who is "
+                        + "calling, not pass a literal boolean -- block: " + handlerBlock);
+    }
+
+    // --- who may see fleet_list's leads and members arrays ---------------------------------------
+
+    /**
+     * {@link FleetMcp#leadsVisibleTo} is the whole policy decision for {@code fleet_list}'s
+     * {@code leads} array: visible to exactly the roles that may {@code SEND} to a lead -- the
+     * primary, an architect, and a collaborator -- never a worker, which holds {@code READ} but
+     * can never {@code SEND} at all, and never an anonymous caller.
+     */
+    @Test
+    void primaryArchitectAndCollaboratorMaySeeTheLeadsArray() {
+        assertTrue(FleetMcp.leadsVisibleTo(PRIMARY), "the primary must see the leads array");
+        assertTrue(FleetMcp.leadsVisibleTo(ARCH_DESIGN), "an architect must see the leads array");
+        assertTrue(FleetMcp.leadsVisibleTo(COLLABORATOR),
+                "a collaborator may SEND to a lead, so it must see the leads array to learn where");
+        assertFalse(FleetMcp.leadsVisibleTo(WORKER_A),
+                "a worker holds READ but can never SEND, so it must not see the leads array");
+        assertFalse(FleetMcp.leadsVisibleTo(ANON), "authenticated as nothing must not see it either");
+    }
+
+    /**
+     * As {@link #primaryArchitectAndCollaboratorMaySeeTheLeadsArray}, for the {@code members}
+     * array -- but a collaborator may {@code SEND} only to a lead or another collaborator, never
+     * to a spawned member, so it must not see this one.
+     */
+    @Test
+    void onlyPrimaryAndArchitectMaySeeTheMembersArray() {
+        assertTrue(FleetMcp.membersVisibleTo(PRIMARY), "the primary must see the members array");
+        assertTrue(FleetMcp.membersVisibleTo(ARCH_DESIGN), "an architect must see the members array");
+        assertFalse(FleetMcp.membersVisibleTo(WORKER_A),
+                "a worker holds READ but can never SEND, so it must not see the members array");
+        assertFalse(FleetMcp.membersVisibleTo(COLLABORATOR),
+                "a collaborator may SEND to a lead, never to a spawned member, so it must not see the members array");
+        assertFalse(FleetMcp.membersVisibleTo(ANON), "authenticated as nothing must not see it either");
+    }
+
+    // --- who may see fleet_list's panes array ----------------------------------------------------
+
+    /**
+     * {@link FleetMcp#panesVisibleTo} is the whole policy decision for {@code fleet_list}'s
+     * {@code panes} array: visible to every role that may {@code SEND} to some other pane -- the
+     * primary, an architect, a collaborator, and an observer (to another observer pane only, with
+     * its row filtered and reduced -- see {@code listFleet}) -- never a worker, never an
+     * anonymous caller.
+     */
+    @Test
+    void onlyPrimaryArchitectCollaboratorAndObserverMaySeeThePanesArray() {
+        assertTrue(FleetMcp.panesVisibleTo(PRIMARY), "the primary must see the panes array");
+        assertTrue(FleetMcp.panesVisibleTo(ARCH_DESIGN), "an architect must see the panes array");
+        assertTrue(FleetMcp.panesVisibleTo(COLLABORATOR), "a collaborator must see its own peer roster");
+        assertFalse(FleetMcp.panesVisibleTo(WORKER_A),
+                "a worker holds READ but can never SEND, so it must not see the panes array");
+        assertTrue(FleetMcp.panesVisibleTo(Principal.observer("term_obs", 700)),
+                "an observer holds SEND to another observer pane, so it must see the (filtered, "
+                        + "reduced) panes array");
+        assertFalse(FleetMcp.panesVisibleTo(ANON), "authenticated as nothing must not see it either");
+    }
+
+    /**
+     * Same reasoning as {@link #theFleetListHandlerActuallyConsultsCoordinatorVisibleTo}: the
+     * predicate above can be perfectly correct while the one production call site never asks it.
+     * This reads {@code FleetMcp.java}'s own source and asserts the {@code fleet_list} handler's
+     * {@code listFleet(...)} call asks {@code leadsVisibleTo(principal(exchange))} for the leads
+     * visibility flag, rather than a literal boolean.
+     */
+    @Test
+    void theFleetListHandlerActuallyConsultsLeadsVisibleTo() throws Exception {
+        String source = Files.readString(MCP_SOURCE);
+
+        int start = source.indexOf("listHandler =");
+        assertTrue(start >= 0, "could not find the fleet_list handler (listHandler) in " + MCP_SOURCE
+                + " -- the scrape has stopped matching, fix the anchor before trusting this test");
+        int end = source.indexOf("stopHandler =", start);
+        assertTrue(end > start, "could not find the handler declared after listHandler to bound the scrape");
+        String handlerBlock = source.substring(start, end);
+
+        // CONTROL: the block we scraped really does contain a call to listFleet(...) -- if this
+        // fails, the anchors above moved and the assertion below would otherwise pass on nothing.
+        assertTrue(handlerBlock.contains("listFleet("),
+                "control failed: the scraped listHandler block contains no listFleet( call at all -- "
+                        + "the anchors have drifted, this test is not testing what it claims to");
+
+        assertTrue(handlerBlock.contains("leadsVisibleTo(principal(exchange))"),
+                "the fleet_list handler must ask leadsVisibleTo(principal(exchange)) who is calling, "
+                        + "not pass a literal boolean -- block: " + handlerBlock);
+    }
+
+    /** As {@link #theFleetListHandlerActuallyConsultsLeadsVisibleTo}, for {@code membersVisibleTo}. */
+    @Test
+    void theFleetListHandlerActuallyConsultsMembersVisibleTo() throws Exception {
+        String source = Files.readString(MCP_SOURCE);
+
+        int start = source.indexOf("listHandler =");
+        assertTrue(start >= 0, "could not find the fleet_list handler (listHandler) in " + MCP_SOURCE
+                + " -- the scrape has stopped matching, fix the anchor before trusting this test");
+        int end = source.indexOf("stopHandler =", start);
+        assertTrue(end > start, "could not find the handler declared after listHandler to bound the scrape");
+        String handlerBlock = source.substring(start, end);
+
+        assertTrue(handlerBlock.contains("listFleet("),
+                "control failed: the scraped listHandler block contains no listFleet( call at all -- "
+                        + "the anchors have drifted, this test is not testing what it claims to");
+
+        assertTrue(handlerBlock.contains("membersVisibleTo(principal(exchange))"),
+                "the fleet_list handler must ask membersVisibleTo(principal(exchange)) who is calling, "
+                        + "not pass a literal boolean -- block: " + handlerBlock);
+    }
+
+    /**
+     * {@code fleet_poll{ticket}} must thread the calling connection's owner key into
+     * {@link MessageService#poll(String, String)}, so a worker cannot read a ticket a different
+     * session created.
+     */
+    @Test
+    void theFleetPollHandlerActuallyThreadsCallerOwnerIntoPoll() throws Exception {
+        String source = Files.readString(MCP_SOURCE);
+
+        int start = source.indexOf("pollHandler =");
+        assertTrue(start >= 0, "could not find the fleet_poll handler (pollHandler) in " + MCP_SOURCE
+                + " -- the scrape has stopped matching, fix the anchor before trusting this test");
+        int end = source.indexOf("ackHandler =", start);
+        assertTrue(end > start, "could not find the handler declared after pollHandler to bound the scrape");
+        String handlerBlock = source.substring(start, end);
+
+        // CONTROL: the block we scraped really does call poll(...) -- if this fails, the anchors
+        // above moved and the assertions below would otherwise pass on nothing.
+        assertTrue(handlerBlock.contains("poll(messages,"),
+                "control failed: the scraped pollHandler block contains no poll(messages, ...) call "
+                        + "at all -- the anchors have drifted, this test is not testing what it claims to");
+
+        assertTrue(handlerBlock.contains("principal(exchange).ownerKey()"),
+                "the fleet_poll handler must thread principal(exchange).ownerKey() into poll(...), not omit "
+                        + "it or pass a literal null -- block: " + handlerBlock);
+    }
+
+    /**
+     * {@code fleet_status} must thread the calling connection's owner key into
+     * {@link FleetMcp#status(MessageService, String, String)}, so a caller that did not create a
+     * worker's open delegation cannot read its pending question through the status handler either.
+     */
+    @Test
+    void theFleetStatusHandlerActuallyThreadsCallerOwnerIntoStatus() throws Exception {
+        String source = Files.readString(MCP_SOURCE);
+
+        int start = source.indexOf("statusHandler =");
+        assertTrue(start >= 0, "could not find the fleet_status handler (statusHandler) in " + MCP_SOURCE
+                + " -- the scrape has stopped matching, fix the anchor before trusting this test");
+        int end = source.indexOf("pollHandler =", start);
+        assertTrue(end > start, "could not find the handler declared after statusHandler to bound the scrape");
+        String handlerBlock = source.substring(start, end);
+
+        // CONTROL: the block we scraped really does call status(...) -- if this fails, the anchors
+        // above moved and the assertions below would otherwise pass on nothing.
+        assertTrue(handlerBlock.contains("status(messages,"),
+                "control failed: the scraped statusHandler block contains no status(messages, ...) "
+                        + "call at all -- the anchors have drifted, this test is not testing what it claims to");
+
+        assertTrue(handlerBlock.contains("principal(exchange).ownerKey()"),
+                "the fleet_status handler must thread principal(exchange).ownerKey() into status(...), not "
+                        + "omit it or pass a literal null -- block: " + handlerBlock);
+    }
+
     // --- which action each tool hands the gate (fleetd #272) ------------------------------------
 
     /**
@@ -297,35 +647,53 @@ class FleetMcpAuthzTest {
      * the whole time the defect was live -- the table was right, the action fed to it was wrong.
      */
     @Test
-    void pollingByTargetIsADrainAndPollingByTicketIsARead() {
+    void pollingByTargetIsADrainAndPollingByTicketIsATaskRead() {
         assertEquals(Authz.Action.DRAIN, FleetMcp.pollAction("term_b", null),
                 "poll by target removes the replies — that is a drain, not an observation");
-        assertEquals(Authz.Action.READ, FleetMcp.pollAction(null, null),
-                "poll by ticket changes nothing");
-        assertEquals(Authz.Action.READ, FleetMcp.pollAction("  ", null),
+        assertEquals(Authz.Action.TASK_READ, FleetMcp.pollAction(null, null),
+                "poll by ticket changes nothing, but is not the roster-only READ action");
+        assertEquals(Authz.Action.TASK_READ, FleetMcp.pollAction("  ", null),
                 "a blank target is an absent target");
     }
 
     /**
      * fleetd #421: a coordId branch is a THIRD operation behind fleet_poll's one name, and it must
-     * map to {@link Authz.Action#COORD_READ} — never {@link Authz.Action#READ}, even though this
-     * branch also consumes nothing. READ's grant is open to every authenticated role on the premise
-     * that the roster carries no secrets; a lead-to-lead body is not the roster, so folding this
-     * branch into READ would let any worker read every peer lead's mail in full. coordId also takes
-     * priority over target when both happen to be present — it addresses a different inbox entirely.
+     * map to {@link Authz.Action#COORD_READ} — never {@link Authz.Action#READ} or {@link
+     * Authz.Action#TASK_READ}, even though this branch also consumes nothing. A lead-to-lead body
+     * is not the roster and not a ticket/status read, so folding this branch into either would let
+     * any worker or architect read every peer lead's mail in full. coordId also takes priority over
+     * target when both happen to be present — it addresses a different inbox entirely.
      */
     @Test
-    void pollingByCoordIdIsACoordReadNeverAPlainRead() {
+    void pollingByCoordIdIsACoordReadNeverAPlainOrTaskRead() {
         assertEquals(Authz.Action.COORD_READ, FleetMcp.pollAction(null, "mac-opus"),
-                "reading held peer mail must not be mapped to the everyone-readable READ action");
+                "reading held peer mail must not be mapped to a widely-readable action");
         assertEquals(Authz.Action.COORD_READ, FleetMcp.pollAction("  ", "mac-opus"),
                 "a blank target must not fall through to READ/DRAIN when coordId is present");
-        assertEquals(Authz.Action.READ, FleetMcp.pollAction(null, "  "),
+        assertEquals(Authz.Action.TASK_READ, FleetMcp.pollAction(null, "  "),
                 "a blank coordId is an absent coordId, same as target/ticket");
         assertEquals(Authz.Action.COORD_READ, FleetMcp.pollAction("term_b", "mac-opus"),
                 "coordId takes priority over target — this is a different inbox, not a drain");
     }
 
+    /**
+     * {@code fleet_send} is three call shapes behind one tool name, exactly as {@code fleet_poll}
+     * is (fleetd #669 Unit A). {@link FleetMcp#sendAction} picks the action from the arguments, not
+     * the handler, for the same reason {@link FleetMcp#pollAction} does: a test can assert the
+     * mapping the handler actually uses.
+     */
+    @Test
+    void sendMapsToThreeDifferentActionsByItsArguments() {
+        assertEquals(Authz.Action.SEND, FleetMcp.sendAction(null, null),
+                "a plain delivery, with neither coordId nor turnId, is a local SEND");
+        assertEquals(Authz.Action.COORD_SEND, FleetMcp.sendAction("mac-opus", null),
+                "coordId addresses a peer lead over the coordination broker");
+        assertEquals(Authz.Action.ANSWER, FleetMcp.sendAction(null, "turn-1"),
+                "turnId resolves a worker's blocked question");
+        assertEquals(Authz.Action.COORD_SEND, FleetMcp.sendAction("mac-opus", "turn-1"),
+                "coordId takes priority over turnId, mirroring sendToLead's own mutual-exclusion check");
+    }
+
     @Test
     void everyRegisteredToolHasItsHandlerActionPinned() {
         // fleetd #469: this used to scrape FleetMcp.java's tool("…") calls for the registered set —
@@ -344,16 +712,22 @@ class FleetMcpAuthzTest {
                 () -> tool + " is registered but has no pinned authorization action"));
 
         assertEquals(Authz.Action.SEND, FleetMcp.toolAction("fleet_send", Map.of()));
+        assertEquals(Authz.Action.SEND,
+                FleetMcp.toolAction("fleet_send", Map.of("sessionId", "term_a", "content", "hi")));
+        assertEquals(Authz.Action.COORD_SEND,
+                FleetMcp.toolAction("fleet_send", Map.of("coordId", "mac-opus", "content", "hi")));
+        assertEquals(Authz.Action.ANSWER,
+                FleetMcp.toolAction("fleet_send", Map.of("turnId", "turn-1", "content", "hi")));
         assertEquals(Authz.Action.REPLY, FleetMcp.toolAction("fleet_reply", Map.of()));
         assertEquals(Authz.Action.ASK, FleetMcp.toolAction("fleet_ask", Map.of()));
-        assertEquals(Authz.Action.READ, FleetMcp.toolAction("fleet_status", Map.of()));
+        assertEquals(Authz.Action.TASK_READ, FleetMcp.toolAction("fleet_status", Map.of()));
         assertEquals(Authz.Action.DRAIN, FleetMcp.toolAction("fleet_ack", Map.of()));
         assertEquals(Authz.Action.SPAWN, FleetMcp.toolAction("fleet_spawn", Map.of()));
         assertEquals(Authz.Action.READ, FleetMcp.toolAction("fleet_list", Map.of()));
         assertEquals(Authz.Action.STOP, FleetMcp.toolAction("fleet_stop", Map.of()));
         assertEquals(Authz.Action.READ, FleetMcp.toolAction("fleet_profiles", Map.of()));
         assertEquals(Authz.Action.READ, FleetMcp.toolAction("fleet_whoami", Map.of()));
-        assertEquals(Authz.Action.READ, FleetMcp.toolAction("fleet_poll", Map.of("ticket", "task")));
+        assertEquals(Authz.Action.TASK_READ, FleetMcp.toolAction("fleet_poll", Map.of("ticket", "task")));
         assertEquals(Authz.Action.DRAIN, FleetMcp.toolAction("fleet_poll", Map.of("target", "term_b")));
         assertEquals(Authz.Action.COORD_READ,
                 FleetMcp.toolAction("fleet_poll", Map.of("coordId", "mac-opus")));

fleetd/src/test/java/dev/ltms/fleet/mcp/FleetMcpHandoverTest.java

@@ -11,6 +11,7 @@ import dev.ltms.fleet.herdr.FakeHerdr;
 import dev.ltms.fleet.herdr.PaneLocator;
 import dev.ltms.fleet.herdr.WorkspaceControl;
 import dev.ltms.fleet.inject.Injector;
+import dev.ltms.fleet.lead.LeadLauncher;
 import dev.ltms.fleet.lead.LeadRollover;
 import dev.ltms.fleet.member.ClaudeCodeLauncher;
 import dev.ltms.fleet.msg.InMemoryReplyInbox;
@@ -24,6 +25,8 @@ import org.junit.jupiter.api.DisplayName;
 import org.junit.jupiter.api.Test;
 import org.junit.jupiter.api.io.TempDir;
 
+import java.io.IOException;
+import java.io.UncheckedIOException;
 import java.nio.file.Files;
 import java.nio.file.Path;
 import java.util.List;
@@ -36,14 +39,14 @@ import static org.junit.jupiter.api.Assertions.*;
  * fleetd #480 Unit C — the {@code fleet_handover} MCP tool, the surface that finally calls
  * {@link LeadRollover#open}/{@link LeadRollover#confirm}/{@link LeadRollover#cancel}.
  *
- * <p>Uses {@link LeadRollover}'s PUBLIC constructor (real wall clock, real 250ms settle poll, a
- * real virtual-thread continuation runner) rather than its package-private test constructor —
- * this test lives in {@code dev.ltms.fleet.mcp}, not {@code dev.ltms.fleet.lead}, and does not
- * need to control the post-{@code confirm()} continuation's timing: it only asserts the
- * SYNCHRONOUS return value of {@code open}/{@code confirm}/{@code cancel}, which is exactly what
- * {@code FleetMcp.handover} forwards to the client. {@code turnSettleSeconds}/{@code
- * clearSettleSeconds} are kept at 1s so a confirmed request's background continuation (which this
- * class does not wait on or assert against) gives up quickly rather than polling for 20s on a
+ * <p>Uses {@link LeadRollover}'s PUBLIC constructor (real wall clock, real 250ms poll, a real
+ * virtual-thread continuation runner) rather than its package-private test constructor — this
+ * test lives in {@code dev.ltms.fleet.mcp}, not {@code dev.ltms.fleet.lead}, and does not need to
+ * control the post-{@code confirm()} continuation's timing: it only asserts the SYNCHRONOUS return
+ * value of {@code open}/{@code confirm}/{@code cancel}, which is exactly what {@code
+ * FleetMcp.handover} forwards to the client. {@code turnSettleSeconds}/{@code
+ * relaunchReadySeconds} are kept at 1s so a confirmed request's background continuation (which
+ * this class does not wait on or assert against) gives up quickly rather than polling for 20s on a
  * daemon virtual thread.
  */
 class FleetMcpHandoverTest {
@@ -53,9 +56,14 @@ class FleetMcpHandoverTest {
 
     private static final String LEAD = "term_lead";
     private static final String OTHER_LEAD = "term_other_lead";
+    private static final String LEAD_OWNER = "leader:lead";
 
     private final FakeHerdr herdr = new FakeHerdr();
     private final AgentControl agents = new AgentControl(herdr);
+    /** Fed to every direct {@code FleetMcp.handover} call below — none of this class's own tests
+     * exercise ticket/ask ownership, so a single instance with no delegations is enough. */
+    private final MessageService messages = new MessageService(agents, new Injector(agents),
+            new Rendezvous(), new InMemoryReplyInbox());
     private FleetMcp mcp;
 
     @AfterEach
@@ -67,10 +75,26 @@ class FleetMcpHandoverTest {
         return new FleetConfig.LeadRollover(handoverPath, false, 3600, 1, 1, "read the handover file");
     }
 
+    private static FleetConfig minimalFleetConfig() {
+        try {
+            Path yaml = Files.createTempFile("fleet-mcp-handover-test", ".yaml");
+            Files.writeString(yaml, "bind:\n  port: 8080\n");
+            return FleetConfig.load(yaml);
+        } catch (IOException e) {
+            throw new UncheckedIOException(e);
+        }
+    }
+
     private LeadRollover newRollover(String handoverPath) {
         // Every handoverPath this test class uses comes from tmp.resolve(...), which is already
         // absolute, so the workspace lookup is never actually consulted — a no-op lookup is enough.
-        return new LeadRollover(agents, () -> cfg(handoverPath), _ -> null);
+        // None of this class's tests reach the recognition-wait or the relaunch call, so the
+        // launcher's own functional correctness is irrelevant here — any constructed instance,
+        // backed by the same fake herdr, is enough.
+        WorkspaceControl spaces = new WorkspaceControl(herdr);
+        LeadLauncher launcher = new LeadLauncher(agents, spaces, minimalFleetConfig());
+        return new LeadRollover(agents, spaces, launcher, () -> cfg(handoverPath),
+                _ -> null, _ -> null, Map::of);
     }
 
     /** A fully wired FleetMcp on fakes (mirrors FleetMcpAuthzTest's helper), plus a leadRollover. */
@@ -132,16 +156,16 @@ class FleetMcpHandoverTest {
     @DisplayName("with leadRollover: absent, every action returns a clean NOT_CONFIGURED refusal and never throws")
     void nullLeadRolloverRefusesCleanlyForEveryAction() {
         McpSchema.CallToolResult open = assertDoesNotThrow(
-                () -> FleetMcp.handover(null, LEAD, Map.of("action", "open")));
+                () -> FleetMcp.handover(null, messages, LEAD, LEAD_OWNER, Map.of("action", "open")));
         assertFalse(open.isError(), "a refusal is not a protocol error: " + textOf(open));
         assertTrue(textOf(open).contains("NOT_CONFIGURED"), textOf(open));
 
-        McpSchema.CallToolResult confirm = assertDoesNotThrow(() -> FleetMcp.handover(null, LEAD,
+        McpSchema.CallToolResult confirm = assertDoesNotThrow(() -> FleetMcp.handover(null, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "confirm", "token", "whatever")));
         assertFalse(confirm.isError());
         assertTrue(textOf(confirm).contains("NOT_CONFIGURED"), textOf(confirm));
 
-        McpSchema.CallToolResult cancel = assertDoesNotThrow(() -> FleetMcp.handover(null, LEAD,
+        McpSchema.CallToolResult cancel = assertDoesNotThrow(() -> FleetMcp.handover(null, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "cancel", "token", "whatever")));
         assertFalse(cancel.isError());
         assertTrue(textOf(cancel).contains("NOT_CONFIGURED"), textOf(cancel));
@@ -150,11 +174,11 @@ class FleetMcpHandoverTest {
     @Test
     @DisplayName("a blank/unknown action is a clean tool error, never an exception")
     void unknownActionIsACleanError() {
-        McpSchema.CallToolResult missing = assertDoesNotThrow(() -> FleetMcp.handover(null, LEAD, Map.of()));
+        McpSchema.CallToolResult missing = assertDoesNotThrow(() -> FleetMcp.handover(null, messages, LEAD, LEAD_OWNER, Map.of()));
         assertTrue(missing.isError());
 
         McpSchema.CallToolResult bogus = assertDoesNotThrow(
-                () -> FleetMcp.handover(null, LEAD, Map.of("action", "bogus")));
+                () -> FleetMcp.handover(null, messages, LEAD, LEAD_OWNER, Map.of("action", "bogus")));
         assertTrue(bogus.isError());
     }
 
@@ -210,7 +234,7 @@ class FleetMcpHandoverTest {
         Files.writeString(handover, "not written yet");
         LeadRollover rollover = newRollover(handover.toString());
 
-        McpSchema.CallToolResult openResult = FleetMcp.handover(rollover, LEAD, Map.of("action", "open"));
+        McpSchema.CallToolResult openResult = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER, Map.of("action", "open"));
         assertFalse(openResult.isError(), textOf(openResult));
         String token = extractToken(textOf(openResult));
 
@@ -219,13 +243,13 @@ class FleetMcpHandoverTest {
         Thread.sleep(50);
         Files.writeString(handover, "the real handover content");
 
-        McpSchema.CallToolResult wrongCaller = FleetMcp.handover(rollover, OTHER_LEAD,
+        McpSchema.CallToolResult wrongCaller = FleetMcp.handover(rollover, messages, OTHER_LEAD, "leader:other-lead",
                 Map.of("action", "confirm", "token", token));
         assertFalse(wrongCaller.isError(), "a refusal is a legitimate outcome, not a protocol error");
         assertTrue(textOf(wrongCaller).contains("NOT_YOUR_ROLLOVER"),
                 "a different lead terminal confirming must surface NOT_YOUR_ROLLOVER: " + textOf(wrongCaller));
 
-        McpSchema.CallToolResult confirmed = FleetMcp.handover(rollover, LEAD,
+        McpSchema.CallToolResult confirmed = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "confirm", "token", token));
         assertFalse(confirmed.isError(), textOf(confirmed));
         assertTrue(textOf(confirmed).contains("\"accepted\":true"),
@@ -239,7 +263,7 @@ class FleetMcpHandoverTest {
     void cancelUnknownTokenIsCleanNotAFailure() {
         LeadRollover rollover = newRollover(tmp.resolve("h.md").toString());
 
-        McpSchema.CallToolResult r = FleetMcp.handover(rollover, LEAD,
+        McpSchema.CallToolResult r = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "cancel", "token", "does-not-exist"));
         assertFalse(r.isError());
         assertTrue(textOf(r).contains("\"cancelled\":false"), textOf(r));
@@ -250,9 +274,9 @@ class FleetMcpHandoverTest {
     @DisplayName("cancel on a token actually opened reports cancelled:true")
     void cancelKnownTokenSucceeds() {
         LeadRollover rollover = newRollover(tmp.resolve("h.md").toString());
-        String token = extractToken(textOf(FleetMcp.handover(rollover, LEAD, Map.of("action", "open"))));
+        String token = extractToken(textOf(FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER, Map.of("action", "open"))));
 
-        McpSchema.CallToolResult r = FleetMcp.handover(rollover, LEAD,
+        McpSchema.CallToolResult r = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "cancel", "token", token));
         assertFalse(r.isError());
         assertTrue(textOf(r).contains("\"cancelled\":true"), textOf(r));
@@ -263,7 +287,7 @@ class FleetMcpHandoverTest {
     @Test
     @DisplayName("status on a null LeadRollover is a clean NOT_CONFIGURED refusal, never a throw")
     void statusWithNullLeadRolloverRefusesCleanly() {
-        McpSchema.CallToolResult r = assertDoesNotThrow(() -> FleetMcp.handover(null, LEAD,
+        McpSchema.CallToolResult r = assertDoesNotThrow(() -> FleetMcp.handover(null, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "status", "token", "whatever")));
         assertFalse(r.isError());
         assertTrue(textOf(r).contains("NOT_CONFIGURED"), textOf(r));
@@ -274,7 +298,7 @@ class FleetMcpHandoverTest {
     void statusOnUnknownTokenReportsUnknown() {
         LeadRollover rollover = newRollover(tmp.resolve("h.md").toString());
 
-        McpSchema.CallToolResult r = FleetMcp.handover(rollover, LEAD,
+        McpSchema.CallToolResult r = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "status", "token", "does-not-exist"));
         assertFalse(r.isError());
         assertTrue(textOf(r).contains("\"state\":\"UNKNOWN\""), textOf(r));
@@ -284,9 +308,9 @@ class FleetMcpHandoverTest {
     @DisplayName("status on a token that is still pending (opened, not confirmed) reports PENDING")
     void statusOnPendingTokenReportsPending() {
         LeadRollover rollover = newRollover(tmp.resolve("h.md").toString());
-        String token = extractToken(textOf(FleetMcp.handover(rollover, LEAD, Map.of("action", "open"))));
+        String token = extractToken(textOf(FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER, Map.of("action", "open"))));
 
-        McpSchema.CallToolResult r = FleetMcp.handover(rollover, LEAD,
+        McpSchema.CallToolResult r = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER,
                 Map.of("action", "status", "token", token));
         assertFalse(r.isError());
         assertTrue(textOf(r).contains("\"state\":\"PENDING\""), textOf(r));
@@ -304,5 +328,40 @@ class FleetMcpHandoverTest {
                 "the tool's own description must advertise the 'status' action: " + tool.description());
     }
 
+    // --- unit 5: open() reports outstanding tickets and open asks ------------------------------
+
+    @Test
+    @DisplayName("open() keeps token/handoverPath/requestedAtMillis and reports empty outstanding collections when the caller has nothing")
+    void openReportsEmptyOutstandingCollectionsWhenCallerHasNothing() {
+        LeadRollover rollover = newRollover(tmp.resolve("h.md").toString());
+
+        McpSchema.CallToolResult r = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER,
+                Map.of("action", "open"));
+        assertFalse(r.isError(), textOf(r));
+        String json = textOf(r);
+        assertTrue(json.contains("\"token\":"), json);
+        assertTrue(json.contains("\"handoverPath\":"), json);
+        assertTrue(json.contains("\"requestedAtMillis\":"), json);
+        assertTrue(json.contains("\"outstandingTickets\":[]"),
+                "a caller with nothing gets an empty array, not an absent key: " + json);
+        assertTrue(json.contains("\"openAsks\":[]"),
+                "a caller with nothing gets an empty array, not an absent key: " + json);
+    }
+
+    @Test
+    @DisplayName("open() reports an owned pending ticket with its phase and target")
+    void openReportsAnOwnedPendingTicketWithItsPhase() {
+        LeadRollover rollover = newRollover(tmp.resolve("h.md").toString());
+        String ticket = messages.sendAsync("term_worker", "a task", null, Principal.leader("lead", LEAD, 1));
+
+        McpSchema.CallToolResult r = FleetMcp.handover(rollover, messages, LEAD, LEAD_OWNER,
+                Map.of("action", "open"));
+        assertFalse(r.isError(), textOf(r));
+        String json = textOf(r);
+        assertTrue(json.contains("\"ticket\":\"" + ticket + "\""), json);
+        assertTrue(json.contains("\"phase\":\"PENDING\""), json);
+        assertTrue(json.contains("\"target\":\"term_worker\""), json);
+    }
+
     // --- acceptance 7 (wiring) is covered by FleetdLeadRolloverWiringTest, unchanged -----------
 }

fleetd/src/test/java/dev/ltms/fleet/mcp/FleetMcpLeadContextGaugeWiringTest.java

@@ -121,6 +121,7 @@ class FleetMcpLeadContextGaugeWiringTest {
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.LoopHealthSource.none(), FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(),
                 FleetMcp.LeadSeatSource.none(), contextGauge, leadConfigDirs,
-                Map.of(LEAD_TERMINAL, LEAD_NAME), LEAD_TERMINAL, FleetMcp.CoordinationSource.none(), false);
+                Map.of(LEAD_TERMINAL, LEAD_NAME), LEAD_TERMINAL, Map.of(), false,
+                FleetMcp.CoordinationSource.none(), false, true, true);
     }
 }

fleetd/src/test/java/dev/ltms/fleet/mcp/FleetMcpObserverSendDeliveryTest.java

@@ -0,0 +1,142 @@
+package dev.ltms.fleet.mcp;
+
+import dev.ltms.fleet.auth.CallerResolver;
+import dev.ltms.fleet.auth.MemberRegistry;
+import dev.ltms.fleet.config.FleetConfig;
+import dev.ltms.fleet.guard.SubscriptionGuard;
+import dev.ltms.fleet.herdr.AgentControl;
+import dev.ltms.fleet.herdr.AgentStatus;
+import dev.ltms.fleet.herdr.FakeHerdr;
+import dev.ltms.fleet.herdr.PaneLocator;
+import dev.ltms.fleet.herdr.WorkspaceControl;
+import dev.ltms.fleet.inject.Injector;
+import dev.ltms.fleet.member.ClaudeCodeLauncher;
+import dev.ltms.fleet.msg.InMemoryReplyInbox;
+import dev.ltms.fleet.msg.MessageService;
+import dev.ltms.fleet.msg.Rendezvous;
+import dev.ltms.fleet.session.FakeWorktrees;
+import dev.ltms.fleet.session.SessionManager;
+import io.modelcontextprotocol.client.McpClient;
+import io.modelcontextprotocol.client.McpSyncClient;
+import io.modelcontextprotocol.client.transport.HttpClientStreamableHttpTransport;
+import io.modelcontextprotocol.spec.McpClientTransport;
+import io.modelcontextprotocol.spec.McpSchema;
+import org.eclipse.jetty.server.Server;
+import org.eclipse.jetty.server.ServerConnector;
+import org.eclipse.jetty.servlet.ServletContextHandler;
+import org.eclipse.jetty.servlet.ServletHolder;
+import org.junit.jupiter.api.AfterEach;
+import org.junit.jupiter.api.Test;
+
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * fleetd #743, driven end to end: a real MCP client over a real HTTP transport, resolved by the
+ * real {@link CallerResolver} to {@link dev.ltms.fleet.auth.Role#OBSERVER}, sending to another
+ * unclassified pane. {@link FleetMcpAuthzTest} already proves {@code denyFor} grants this case and
+ * that the handler calls {@code attributeIfObserver}; this test is the one path that also proves
+ * the grant is not dead at a second gate (CB-548's two-gate trap) by driving the real
+ * {@link Injector} to the point of its real herdr call, and reads the exact text the receiving
+ * pane would see.
+ */
+class FleetMcpObserverSendDeliveryTest {
+
+    private static final String TARGET = "term_other_observer";
+
+    private final FakeHerdr herdr = new FakeHerdr();
+    private final AgentControl agents = new AgentControl(herdr);
+    private final Injector injector = new Injector(agents);
+    private final Rendezvous rendezvous = new Rendezvous();
+    private final MessageService messages = new MessageService(agents, injector, rendezvous,
+            new InMemoryReplyInbox());
+    private FleetMcp mcp;
+    private Server server;
+
+    @AfterEach
+    void tearDown() throws Exception {
+        if (server != null) {
+            server.stop();
+        }
+        if (mcp != null) {
+            mcp.close();
+        }
+    }
+
+    @Test
+    void anObserversSendIsAttributedAndReachesTheRealInjector() throws Exception {
+        FleetConfig.Profile cfg = new FleetConfig.Profile(
+                "ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
+                "tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
+        ClaudeCodeLauncher workers = new ClaudeCodeLauncher(agents, new WorkspaceControl(herdr),
+                new SubscriptionGuard(Set.of("gx00.gw")), Map.of(cfg.profile(), cfg), cfg.profile(),
+                _ -> "tok");
+        SessionManager sessions = new SessionManager(workers, new FakeWorktrees());
+        // The fake's pane list carries a second pane, "term_shell", whose shell pid is 9001 and
+        // which hosts no agent -- a herdr-owned pane recognised as no lead, architect, collaborator
+        // or live spawned member, so the real resolver lands it on the observer floor.
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> 9001L);
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
+                Map::of, new MemberRegistry(null));
+
+        mcp = new FleetMcp(messages, workers, sessions, identity, sessions.asPresence(),
+                new PrimaryRegistry(null), callers, FleetMcp.AuthorizationMode.ENFORCED,
+                null, FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
+                FleetMcp.QuarantineSource.none(), null, FleetMcp.OutageSource.none(),
+                FleetMcp.LeadSeatSource.none(), List.of(), null);
+
+        ServletContextHandler handler = new ServletContextHandler();
+        handler.setContextPath("/");
+        handler.addServlet(new ServletHolder(mcp.servlet()), "/mcp");
+        server = new Server(0);
+        server.setHandler(handler);
+        server.start();
+        String baseUrl = "http://127.0.0.1:"
+                + ((ServerConnector) server.getConnectors()[0]).getLocalPort();
+
+        McpSchema.CallToolResult result = sendFleetSend(baseUrl, TARGET, "hi there");
+        assertFalse(result.isError(), "an observer sending to another observer must be accepted: "
+                + textOf(result));
+
+        long waiterDeadline = System.currentTimeMillis() + 3000;
+        while (!rendezvous.isWaiting(TARGET) && System.currentTimeMillis() < waiterDeadline) {
+            Thread.sleep(5);
+        }
+        assertTrue(rendezvous.isWaiting(TARGET), "the async send must have opened its rendezvous waiter");
+
+        injector.onStatus(TARGET, AgentStatus.IDLE); // drives the real delivery attempt to herdr
+
+        long deliveryDeadline = System.currentTimeMillis() + 3000;
+        while (!herdr.called("agent.prompt") && System.currentTimeMillis() < deliveryDeadline) {
+            Thread.sleep(5);
+        }
+        assertTrue(herdr.called("agent.prompt"), "the delivery attempt must have reached herdr");
+
+        @SuppressWarnings("unchecked")
+        Map<String, Object> params = (Map<String, Object>) herdr.lastCall("agent.prompt").params();
+        assertEquals("[fleet_send from observer term_shell]\nhi there", params.get("text"),
+                "the receiving pane must see the sender's own daemon-resolved terminal, never a raw "
+                        + "echo of the content and never a client-supplied name");
+    }
+
+    private static McpSchema.CallToolResult sendFleetSend(String baseUrl, String target, String content) {
+        McpClientTransport transport = HttpClientStreamableHttpTransport.builder(baseUrl)
+                .endpoint("/mcp")
+                .build();
+        try (McpSyncClient client = McpClient.sync(transport).build()) {
+            client.initialize();
+            return client.callTool(McpSchema.CallToolRequest.builder("fleet_send")
+                    .arguments(Map.of("sessionId", target, "content", content, "wait", false))
+                    .build());
+        }
+    }
+
+    private static String textOf(McpSchema.CallToolResult r) {
+        return ((McpSchema.TextContent) r.content().getFirst()).text();
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/mcp/FleetMcpTest.java

@@ -1,5 +1,6 @@
 package dev.ltms.fleet.mcp;
 
+import dev.ltms.fleet.Fleetd;
 import dev.ltms.fleet.auth.CallerResolver;
 import dev.ltms.fleet.auth.MemberRegistry;
 import dev.ltms.fleet.auth.Principal;
@@ -10,6 +11,7 @@ import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
 import dev.ltms.fleet.herdr.FakeHerdr;
 import dev.ltms.fleet.inject.LoopWatchdog;
+import dev.ltms.fleet.lead.LeadContextGauge;
 import dev.ltms.fleet.herdr.PaneLocator;
 import dev.ltms.fleet.herdr.WorkspaceControl;
 import dev.ltms.fleet.inject.Injector;
@@ -76,7 +78,7 @@ class FleetMcpTest {
 
     private void assertSendRoundTrips(String target, Set<String> profiles) throws Exception {
         CompletableFuture<McpSchema.CallToolResult> send = CompletableFuture.supplyAsync(
-                () -> FleetMcp.send(messages, target, "hi", 4000L, null, profiles));
+                () -> FleetMcp.send(messages, target, "hi", 4000L, null, profiles, null));
         long deadline = System.currentTimeMillis() + 3000;
         while (!rendezvous.isWaiting(target) && System.currentTimeMillis() < deadline) {
             Thread.sleep(5);
@@ -102,7 +104,7 @@ class FleetMcpTest {
     void sendThenReplyRoundTrips() throws Exception {
         // fleet_send blocks; fleet_reply resolves it with the worker's structured answer.
         CompletableFuture<McpSchema.CallToolResult> send = CompletableFuture.supplyAsync(
-                () -> FleetMcp.send(messages, "term_a", "review this", 4000L, null, Set.of()));
+                () -> FleetMcp.send(messages, "term_a", "review this", 4000L, null, Set.of(), null));
 
         // Wait until the send has opened its waiter so the reply resolves it (CB-307: reply now
         // queues in the inbox if no waiter is open, which would break the round-trip).
@@ -180,7 +182,7 @@ class FleetMcpTest {
         String turnId = afterTurnId.substring(0, afterTurnId.indexOf('"'));
 
         CompletableFuture<McpSchema.CallToolResult> answer = CompletableFuture.supplyAsync(
-                () -> FleetMcp.answer(messages, turnId, "config.yaml", 5000L));
+                () -> FleetMcp.answer(messages, turnId, "config.yaml", 5000L, null));
         assertEquals("config.yaml", textOf(ask.get(6, TimeUnit.SECONDS)));
 
         deadline = System.currentTimeMillis() + 3000;
@@ -278,7 +280,7 @@ class FleetMcpTest {
         assertEquals("late reply", messages.drainReplies("term_a").getFirst().content());
 
         CompletableFuture<McpSchema.CallToolResult> answer = CompletableFuture.supplyAsync(
-                () -> FleetMcp.answer(messages, firstTurnId, "config.yaml", 5000L));
+                () -> FleetMcp.answer(messages, firstTurnId, "config.yaml", 5000L, null));
         assertEquals("config.yaml", textOf(ask.get(6, TimeUnit.SECONDS)));
         while (!rendezvous.isWaiting("term_a") && System.currentTimeMillis() < deadline) {
             Thread.sleep(5);
@@ -287,6 +289,91 @@ class FleetMcpTest {
         assertEquals("done", textOf(answer.get(6, TimeUnit.SECONDS)));
     }
 
+    // --- fleetd #715: fleet_send{turnId} is gated on the caller that owns the turn -------------
+
+    /**
+     * A blocking {@code fleet_send} from one caller opens the turn; a {@code fleet_send{turnId}}
+     * from a different caller is refused as an error, and the real owner's answer still succeeds.
+     */
+    @Test
+    void aDifferentCallersMcpAnswerIsRefusedForABlockingSendButTheRealOwnerSucceeds() throws Exception {
+        CompletableFuture<McpSchema.CallToolResult> send = CompletableFuture.supplyAsync(
+                () -> FleetMcp.send(messages, T, "do X", 5000L, null, Set.of(), "term_owner"));
+        long deadline = System.currentTimeMillis() + 3000;
+        while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
+            Thread.sleep(5);
+        }
+        assertTrue(rendezvous.isWaiting(T));
+
+        CompletableFuture<McpSchema.CallToolResult> ask = CompletableFuture.supplyAsync(
+                () -> FleetMcp.ask(messages, T, "which config?", 5000L));
+        McpSchema.CallToolResult question = send.get(5, TimeUnit.SECONDS);
+        assertTrue(textOf(question).contains("[question]"), textOf(question));
+        String questionText = textOf(question);
+        String afterTurnId = questionText.substring(questionText.indexOf("turnId=\"") + "turnId=\"".length());
+        String turnId = afterTurnId.substring(0, afterTurnId.indexOf('"'));
+
+        McpSchema.CallToolResult hijacked = FleetMcp.answer(messages, turnId, "evil.yaml", 500L, "term_attacker");
+        assertTrue(hijacked.isError(), "a caller that did not open this turn must get an error, not an answer");
+        assertFalse(ask.isDone(), "a refused answer must not resolve the worker's blocked fleet_ask");
+
+        CompletableFuture<McpSchema.CallToolResult> answer = CompletableFuture.supplyAsync(
+                () -> FleetMcp.answer(messages, turnId, "config.yaml", 5000L, "term_owner"));
+        assertEquals("config.yaml", textOf(ask.get(5, TimeUnit.SECONDS)));
+        deadline = System.currentTimeMillis() + 3000;
+        while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
+            Thread.sleep(5);
+        }
+        FleetMcp.reply(messages, T, Role.WORKER, "done");
+        assertEquals("done", textOf(answer.get(5, TimeUnit.SECONDS)));
+    }
+
+    /**
+     * Same hijack and control through the fire-and-poll ({@code sendAsync}) path: the owner comes
+     * from the resolved principal, not from a caller argument threaded through a live call.
+     */
+    @Test
+    void aDifferentCallersMcpAnswerIsRefusedForAnAsyncSendButTheRealOwnerSucceeds() throws Exception {
+        Principal owner = Principal.worker("term_owner", 1);
+        McpSchema.CallToolResult accepted =
+                FleetMcp.sendAsync(messages, T, "do it", null, Set.of(), owner);
+        String ticket = textOf(accepted).substring(textOf(accepted).indexOf("ticket=") + "ticket=".length()).trim();
+
+        long deadline = System.currentTimeMillis() + 3000;
+        while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
+            Thread.sleep(5);
+        }
+        assertTrue(rendezvous.isWaiting(T));
+
+        CompletableFuture<McpSchema.CallToolResult> ask = CompletableFuture.supplyAsync(
+                () -> FleetMcp.ask(messages, T, "which config?", 5000L));
+        MessageService.TaskView asking = messages.poll(ticket);
+        deadline = System.currentTimeMillis() + 3000;
+        while (asking.phase() != MessageService.Phase.ASKING && System.currentTimeMillis() < deadline) {
+            Thread.sleep(5);
+            asking = messages.poll(ticket);
+        }
+        assertEquals(MessageService.Phase.ASKING, asking.phase());
+        String turnId = asking.turnId();
+
+        McpSchema.CallToolResult hijacked = FleetMcp.answer(messages, turnId, "evil.yaml", 500L,
+                "worker:term_attacker");
+        assertTrue(hijacked.isError(), "a caller that did not create this delegation must get an error");
+        assertFalse(ask.isDone(), "a refused answer must not resolve the worker's blocked fleet_ask");
+        assertEquals(MessageService.Phase.ASKING, messages.poll(ticket).phase(),
+                "a refused answer must not advance the async ticket's phase");
+
+        CompletableFuture<McpSchema.CallToolResult> answer = CompletableFuture.supplyAsync(
+                () -> FleetMcp.answer(messages, turnId, "config.yaml", 5000L, owner.ownerKey()));
+        assertEquals("config.yaml", textOf(ask.get(5, TimeUnit.SECONDS)));
+        deadline = System.currentTimeMillis() + 3000;
+        while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
+            Thread.sleep(5);
+        }
+        FleetMcp.reply(messages, T, Role.WORKER, "done");
+        assertEquals("done", textOf(answer.get(5, TimeUnit.SECONDS)));
+    }
+
     @Test
     void pollUnknownTicketIsAnError() {
         McpSchema.CallToolResult res = FleetMcp.poll(messages, "task-999", null);
@@ -296,7 +383,7 @@ class FleetMcpTest {
 
     @Test
     void sendTimesOutWithAWorkingNote() {
-        McpSchema.CallToolResult res = FleetMcp.send(messages, "term_a", "hi", 120L, null, Set.of());
+        McpSchema.CallToolResult res = FleetMcp.send(messages, "term_a", "hi", 120L, null, Set.of(), null);
         assertNotEquals(Boolean.TRUE, res.isError(), "a timeout is informational, not a tool error");
         assertTrue(textOf(res).contains("no reply"), "got: " + textOf(res));
     }
@@ -312,7 +399,7 @@ class FleetMcpTest {
     void sendTimesOutWithAnUnconfirmedNoteNotARetryInvitation() throws Exception {
         herdr.agentSendFailsWith("send_failed");
         CompletableFuture<McpSchema.CallToolResult> send = CompletableFuture.supplyAsync(
-                () -> FleetMcp.send(messages, T, "hi", 150L, null, Set.of()));
+                () -> FleetMcp.send(messages, T, "hi", 150L, null, Set.of(), null));
         long deadline = System.currentTimeMillis() + 2000;
         while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
             //noinspection BusyWait
@@ -332,13 +419,13 @@ class FleetMcpTest {
 
     @Test
     void sendRejectsMissingArgs() {
-        assertTrue(FleetMcp.send(messages, null, "hi", null, null, Set.of()).isError());
-        assertTrue(FleetMcp.send(messages, "term_a", "  ", null, null, Set.of()).isError());
+        assertTrue(FleetMcp.send(messages, null, "hi", null, null, Set.of(), null).isError());
+        assertTrue(FleetMcp.send(messages, "term_a", "  ", null, null, Set.of(), null).isError());
     }
 
     @Test
     void sendRejectsAConfiguredProfileNameBeforeAcceptingIt() {
-        McpSchema.CallToolResult blocking = FleetMcp.send(messages, "sol", "hi", 100L, null, Set.of("sol"));
+        McpSchema.CallToolResult blocking = FleetMcp.send(messages, "sol", "hi", 100L, null, Set.of("sol"), null);
         McpSchema.CallToolResult async = FleetMcp.sendAsync(messages, "sol", "hi", null, Set.of("sol"));
 
         assertTrue(blocking.isError());
@@ -357,7 +444,7 @@ class FleetMcpTest {
         assertSendRoundTrips("term_live_member", profiles);
 
         // A herdr-owned pane outside the bridge roster cannot be classified at accept time.
-        McpSchema.CallToolResult result = FleetMcp.send(messages, "external-pane", "hi", 10L, null, profiles);
+        McpSchema.CallToolResult result = FleetMcp.send(messages, "external-pane", "hi", 10L, null, profiles, null);
         assertFalse(result.isError(), "an unclassified target must not be rejected at acceptance time");
     }
 
@@ -449,7 +536,7 @@ class FleetMcpTest {
     void askThenAnswerRoundTrips() throws Exception {
         // The primary delegates and blocks; wait until its waiter is open before the worker asks.
         CompletableFuture<McpSchema.CallToolResult> send = CompletableFuture.supplyAsync(
-                () -> FleetMcp.send(messages, "term_a", "do X", 5000L, null, Set.of()));
+                () -> FleetMcp.send(messages, "term_a", "do X", 5000L, null, Set.of(), null));
         long deadline = System.currentTimeMillis() + 3000;
         while (!rendezvous.isWaiting("term_a") && System.currentTimeMillis() < deadline) {
             //noinspection BusyWait
@@ -471,7 +558,7 @@ class FleetMcpTest {
 
         // The primary answers via fleet_send(turnId); this blocks again for the worker's reply.
         CompletableFuture<McpSchema.CallToolResult> answer = CompletableFuture.supplyAsync(
-                () -> FleetMcp.answer(messages, turnId, "config.yaml", 5000L));
+                () -> FleetMcp.answer(messages, turnId, "config.yaml", 5000L, null));
 
         // The worker's ask returns the answer — it resumes the same turn.
         assertEquals("config.yaml", textOf(ask.get(6, TimeUnit.SECONDS)));
@@ -497,7 +584,7 @@ class FleetMcpTest {
 
     @Test
     void answerToAStaleTurnIsAnError() {
-        McpSchema.CallToolResult res = FleetMcp.answer(messages, "term_a#999", "too late", 500L);
+        McpSchema.CallToolResult res = FleetMcp.answer(messages, "term_a#999", "too late", 500L, null);
         assertTrue(res.isError());
         assertTrue(textOf(res).contains("no longer open"), textOf(res));
     }
@@ -691,7 +778,7 @@ class FleetMcpTest {
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true);
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true, true, true);
 
         String out = textOf(res);
         // fleetd #361: reports both which coord-id a peer must use to reach ME, and this daemon's
@@ -720,7 +807,7 @@ class FleetMcpTest {
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true);
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true, true, true);
 
         String out = textOf(res);
         assertTrue(out.contains("\"mailbox\":{\"status\":\"unknown\"}"), out);
@@ -779,19 +866,82 @@ class FleetMcpTest {
         SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")));
         FakeLeadChannel channel = new FakeLeadChannel("mac-opus")
                 .withMailbox("mac-opus", LeadChannel.MailboxState.exists("mac-opus", 0, 1));
-        boolean callerIsPrimary = Principal.worker("term_a", 1).isPrimary();
+        Principal worker = Principal.worker("term_a", 1);
 
         McpSchema.CallToolResult res = FleetMcp.listFleet(
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), callerIsPrimary);
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), worker.isPrimary(),
+                FleetMcp.leadsVisibleTo(worker), FleetMcp.membersVisibleTo(worker));
 
         String out = textOf(res);
         assertFalse(out.contains("\"coordinator\""), "a worker must never see the coordinator key at all: " + out);
         assertFalse(out.contains("mac-opus"), "no fragment of the coordinator row may leak either: " + out);
-        assertTrue(out.contains("\"leads\""), "the rest of the result must still be present: " + out);
-        assertTrue(out.contains("\"members\""), out);
+        assertFalse(out.contains("\"leads\""), "a worker must never see the leads key at all: " + out);
+        assertFalse(out.contains("\"members\""), "a worker must never see the members key at all: " + out);
+        assertTrue(out.contains("\"healthCoverage\""), "the rest of the result must still be present: " + out);
+    }
+
+    /**
+     * A worker's result must carry neither the {@code leads} nor the {@code members} key, and no
+     * fragment of either row leaks even though both are fully populated for this call -- a key
+     * check alone would pass on an implementation that still built the rows and only renamed or
+     * nested them.
+     */
+    @Test
+    void listLeaksNoLeadOrMemberRowFragmentToAWorker() {
+        FakeHerdr h = new FakeHerdr();
+        FakeWorktrees worktrees = new FakeWorktrees().withRepoRoot("/repo").withPrefix("/wt");
+        SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), worktrees);
+        MemberSession s = sessions.acquire("ltms-local", null, "/caller/proj", "term_owner",
+                new WorktreeRequest("cb-304", null));
+        Principal worker = Principal.worker("term_a", 1);
+
+        McpSchema.CallToolResult res = FleetMcp.listFleet(
+                workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
+                FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
+                FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
+                Map.of("term_lead_x", "mac-opus"), "", FleetMcp.CoordinationSource.none(), worker.isPrimary(),
+                FleetMcp.leadsVisibleTo(worker), FleetMcp.membersVisibleTo(worker));
+
+        String out = textOf(res);
+        assertFalse(out.contains("\"leads\""), out);
+        assertFalse(out.contains("\"members\""), out);
+        assertFalse(out.contains(s.terminalId()), "no member row fragment may leak to a worker: " + out);
+        assertFalse(out.contains("term_owner"), "no member owner fragment may leak to a worker: " + out);
+        assertFalse(out.contains("mac-opus"), "no lead row fragment may leak to a worker: " + out);
+        assertTrue(out.contains("\"healthCoverage\""), out);
+    }
+
+    /**
+     * A collaborator may {@code SEND} to a lead, so it must see the {@code leads} array -- the
+     * only place {@code fleet_whoami} does not already give it a lead's address. It may never
+     * {@code SEND} to a spawned member, so the {@code members} key must stay absent for it, with
+     * no fragment of a populated member row leaking either.
+     */
+    @Test
+    void listShowsLeadsButNotMembersToACollaborator() {
+        FakeHerdr h = new FakeHerdr();
+        FakeWorktrees worktrees = new FakeWorktrees().withRepoRoot("/repo").withPrefix("/wt");
+        SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), worktrees);
+        MemberSession s = sessions.acquire("ltms-local", null, "/caller/proj", "term_owner",
+                new WorktreeRequest("cb-304", null));
+        Principal collaborator = Principal.collaborator("ops", "term_collab", 600);
+
+        McpSchema.CallToolResult res = FleetMcp.listFleet(
+                workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
+                FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
+                FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
+                Map.of("term_lead_x", "mac-opus"), "", FleetMcp.CoordinationSource.none(), collaborator.isPrimary(),
+                FleetMcp.leadsVisibleTo(collaborator), FleetMcp.membersVisibleTo(collaborator));
+
+        String out = textOf(res);
+        assertTrue(out.contains("\"leads\""), "a collaborator must see the leads array: " + out);
+        assertTrue(out.contains("mac-opus"), "a collaborator must see the lead's name/address: " + out);
+        assertFalse(out.contains("\"members\""), "a collaborator must never see the members key: " + out);
+        assertFalse(out.contains(s.terminalId()), "no member row fragment may leak to a collaborator: " + out);
+        assertFalse(out.contains("term_owner"), "no member owner fragment may leak to a collaborator: " + out);
         assertTrue(out.contains("\"healthCoverage\""), out);
     }
 
@@ -807,16 +957,19 @@ class FleetMcpTest {
         SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")));
         FakeLeadChannel channel = new FakeLeadChannel("mac-opus")
                 .withMailbox("mac-opus", LeadChannel.MailboxState.exists("mac-opus", 0, 1));
-        boolean callerIsPrimary = Principal.architect("lead-designer", "term_design", 400).isPrimary();
+        Principal architect = Principal.architect("lead-designer", "term_design", 400);
 
         McpSchema.CallToolResult res = FleetMcp.listFleet(
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), callerIsPrimary);
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), architect.isPrimary(),
+                FleetMcp.leadsVisibleTo(architect), FleetMcp.membersVisibleTo(architect));
 
         String out = textOf(res);
         assertFalse(out.contains("\"coordinator\""), "an architect must never see the coordinator key either: " + out);
+        assertTrue(out.contains("\"leads\""), "an architect must still see the leads array: " + out);
+        assertTrue(out.contains("\"members\""), "an architect must still see the members array: " + out);
     }
 
     /**
@@ -841,7 +994,7 @@ class FleetMcpTest {
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true));
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true, true, true));
 
         assertTrue(gatedAsPrimary.contains("\"coordinator\""), gatedAsPrimary);
         assertTrue(gatedAsPrimary.contains("\"selfId\":\"mac-opus\""), gatedAsPrimary);
@@ -850,6 +1003,8 @@ class FleetMcpTest {
         assertTrue(gatedAsPrimary.contains("\"heldDurable\""), gatedAsPrimary);
         assertTrue(gatedAsPrimary.contains("\"held\""), gatedAsPrimary);
         assertTrue(gatedAsPrimary.contains("\"peers\""), gatedAsPrimary);
+        assertTrue(gatedAsPrimary.contains("\"leads\""), "the primary must still see the leads array: " + gatedAsPrimary);
+        assertTrue(gatedAsPrimary.contains("\"members\""), "the primary must still see the members array: " + gatedAsPrimary);
     }
 
     @Test
@@ -868,7 +1023,7 @@ class FleetMcpTest {
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true);
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true, true, true);
 
         String out = textOf(res);
         assertTrue(out.contains("\"msgId\":\"m1\""), out);
@@ -879,6 +1034,330 @@ class FleetMcpTest {
         assertTrue(out.contains("x".repeat(80) + "…"), "expected an 80-char preview with an ellipsis: " + out);
     }
 
+    // --- fleetd #703: fleet_list's collaborators array -------------------------------------------
+
+    /** Calls the canonical {@code listFleet} overload directly, so a test can set the collaborators
+     * payload and its visibility independently of a real {@code Principal} / MCP exchange. */
+    private static McpSchema.CallToolResult listFleetWithCollaborators(FakeHerdr h,
+            Map<String, String> collaborators, boolean collaboratorsVisible) {
+        SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")));
+        return FleetMcp.listFleet(
+                workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
+                FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
+                FleetMcp.LoopHealthSource.none(), FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(),
+                FleetMcp.LeadSeatSource.none(), new LeadContextGauge(), FleetMcp.LeadConfigDirSource.none(),
+                Map.of(), "", collaborators, collaboratorsVisible,
+                FleetMcp.CoordinationSource.none(), false, true, true);
+    }
+
+    /**
+     * fleetd #703 acceptance A: a visible caller with one configured collaborator gets a
+     * {@code collaborators} row whose key ({@code CallerResolver.collaborators()}'s
+     * {@code terminal_id -> name} entry) lands as that row's {@code sessionId}, and {@code leads}/
+     * {@code members} are unaffected by the new key.
+     */
+    @Test
+    void listReportsACollaboratorsRowKeyedByTheCollaboratorsTerminalId() {
+        FakeHerdr h = new FakeHerdr();
+        String out = textOf(listFleetWithCollaborators(h, Map.of("term_collab", "ops"), true));
+
+        assertTrue(out.contains("\"collaborators\":["), out);
+        assertTrue(out.contains("\"name\":\"ops\""), out);
+        assertTrue(out.contains("\"sessionId\":\"term_collab\""), out);
+        assertTrue(out.contains("\"leads\":[]"), out);
+        assertTrue(out.contains("\"members\":[]"), out);
+    }
+
+    /**
+     * fleetd #703 acceptance A, control half: with no collaborator configured, the key is absent
+     * (caller cannot see it) or an empty array (caller can), and {@code leads}/{@code members} are
+     * unchanged either way.
+     */
+    @Test
+    void listOmitsOrEmptiesCollaboratorsWhenNoneAreConfigured() {
+        FakeHerdr h = new FakeHerdr();
+
+        String visibleButEmpty = textOf(listFleetWithCollaborators(h, Map.of(), true));
+        assertTrue(visibleButEmpty.contains("\"collaborators\":[]"), visibleButEmpty);
+        assertTrue(visibleButEmpty.contains("\"leads\":[]"), visibleButEmpty);
+        assertTrue(visibleButEmpty.contains("\"members\":[]"), visibleButEmpty);
+
+        String notVisible = textOf(listFleetWithCollaborators(h, Map.of(), false));
+        assertFalse(notVisible.contains("\"collaborators\""), notVisible);
+        assertTrue(notVisible.contains("\"leads\":[]"), notVisible);
+        assertTrue(notVisible.contains("\"members\":[]"), notVisible);
+    }
+
+    /**
+     * fleetd #703 acceptance B: a worker must not see the {@code collaborators} array at all, while
+     * an architect -- a role that also holds READ, same as a worker -- does see it. Both halves are
+     * asserted: a test that only checked the worker-hidden half would pass even if the feature were
+     * never wired up for anyone.
+     */
+    @Test
+    void listOmitsCollaboratorsForAWorkerAndIncludesThemForAnArchitect() {
+        FakeHerdr h = new FakeHerdr();
+        Map<String, String> collaborators = Map.of("term_collab", "ops");
+
+        String asWorker = textOf(listFleetWithCollaborators(h, collaborators,
+                FleetMcp.collaboratorsVisibleTo(Principal.worker("term_w", 1))));
+        assertFalse(asWorker.contains("\"collaborators\""),
+                "a worker must not see the collaborators array: " + asWorker);
+
+        String asArchitect = textOf(listFleetWithCollaborators(h, collaborators,
+                FleetMcp.collaboratorsVisibleTo(Principal.architect("design", "term_arch", 2))));
+        assertTrue(asArchitect.contains("\"collaborators\":["),
+                "an architect must see the collaborators array: " + asArchitect);
+        assertTrue(asArchitect.contains("\"sessionId\":\"term_collab\""), asArchitect);
+    }
+
+    // --- fleet_list's panes array -----------------------------------------------------------------
+
+    /** Calls the canonical {@code listFleet} overload directly, so a test can set the pane-discovery
+     * payload and its visibility independently of a real {@code Principal} / MCP exchange. */
+    private static McpSchema.CallToolResult listFleetWithPanes(FakeHerdr h, FleetMcp.PaneSource panes,
+            boolean panesVisible) {
+        SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")));
+        return FleetMcp.listFleet(
+                workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
+                FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
+                FleetMcp.LoopHealthSource.none(), FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(),
+                FleetMcp.LeadSeatSource.none(), new LeadContextGauge(), FleetMcp.LeadConfigDirSource.none(),
+                Map.of(), "", Map.of(), false,
+                FleetMcp.CoordinationSource.none(), false, true, true, panes, panesVisible);
+    }
+
+    /**
+     * A pane whose tab herdr reports with a label gets that label and the exact terminal id
+     * {@code fleet_send} takes as a target, carried as {@code sessionId}. fleetd #771: the pane's
+     * workspace carries its herdr space name too, next to {@code workspaceId}.
+     */
+    @Test
+    void listReportsAPaneRowWithItsTabLabelAndSendableSessionId() {
+        FakeHerdr h = new FakeHerdr().withTab("w2", "w2:t7", "trinotes");
+        MemberPresence presence = new MemberPresence();
+        presence.markPresent("term_a");
+        FleetMcp.PaneSource panes = new FleetMcp.PaneSource(
+                () -> new PaneLocator(h).tabLabelsByTabId(),
+                () -> new PaneLocator(h).workspaceLabelsByWorkspaceId(),
+                Fleetd.deliverableTo(presence, Map::of, Map::of), _ -> false, _ -> false);
+
+        String out = textOf(listFleetWithPanes(h, panes, true));
+
+        assertTrue(out.contains("\"panes\":["), out);
+        assertTrue(out.contains("\"sessionId\":\"term_a\""), out);
+        assertTrue(out.contains("\"label\":\"trinotes\""), out);
+        assertTrue(out.contains("\"workspaceLabel\":\"ltms\""),
+                "term_a's agent lives on workspace w2, whose herdr label is \"ltms\": " + out);
+        assertTrue(out.contains("\"deliverable\":true"), out);
+    }
+
+    /**
+     * A pane whose tab carries no label known to herdr still gets a row -- a missing label must
+     * never throw, and must never drop the pane from the array, only report a {@code null} label.
+     * Pairs with a {@code deliverable} false reading when the target is neither present, a lead,
+     * nor a collaborator.
+     */
+    @Test
+    void listReportsAPaneRowWithANullLabelWhenHerdrHasNoneAndNotDeliverable() {
+        FakeHerdr h = new FakeHerdr();
+        FleetMcp.PaneSource panes = new FleetMcp.PaneSource(
+                () -> new PaneLocator(h).tabLabelsByTabId(),
+                () -> new PaneLocator(h).workspaceLabelsByWorkspaceId(),
+                Fleetd.deliverableTo(new MemberPresence(), Map::of, Map::of), _ -> false, _ -> false);
+
+        String out = textOf(listFleetWithPanes(h, panes, true));
+
+        assertTrue(out.contains("\"panes\":["), out);
+        assertTrue(out.contains("\"sessionId\":\"term_a\""), out);
+        assertTrue(out.contains("\"label\":null"), out);
+        assertTrue(out.contains("\"deliverable\":false"), out);
+    }
+
+    /**
+     * fleetd #771: a pane whose agent lives in a workspace that {@code workspace.list} does not
+     * report (an unknown {@code workspaceId}) still gets a row -- the lookup miss must never throw,
+     * and must never drop the pane, only report a {@code null} "workspaceLabel".
+     */
+    @Test
+    void listReportsANullWorkspaceLabelForAnUnknownWorkspaceId() {
+        // withAgent seeds its pane under workspace_id "wQ", which the fake's workspace.list never
+        // reports (only "w1"/"w2") -- modelling a workspace the lookup has no entry for.
+        FakeHerdr h = new FakeHerdr().withAgent("claude-x", "term_unknown_ws", "wQ:p1", "wQ:t1");
+        FleetMcp.PaneSource panes = new FleetMcp.PaneSource(
+                () -> new PaneLocator(h).tabLabelsByTabId(),
+                () -> new PaneLocator(h).workspaceLabelsByWorkspaceId(),
+                _ -> false, _ -> false, _ -> false);
+
+        String out = textOf(listFleetWithPanes(h, panes, true));
+
+        assertTrue(out.contains("\"sessionId\":\"term_unknown_ws\""), out);
+        assertTrue(out.contains("\"workspaceId\":\"wQ\""), out);
+        assertTrue(out.contains("\"workspaceLabel\":null"),
+                "an unknown workspaceId must project a null workspaceLabel, not throw or drop the row: " + out);
+    }
+
+    /** A caller this role may not show the array to gets no {@code panes} key at all. */
+    @Test
+    void listOmitsThePanesArrayWhenTheCallerMayNotSeeIt() {
+        FakeHerdr h = new FakeHerdr();
+
+        String out = textOf(listFleetWithPanes(h, FleetMcp.PaneSource.none(), false));
+
+        assertFalse(out.contains("\"panes\""), out);
+    }
+
+    /**
+     * The tab-label scan behind {@code panes} shares no failure path with the rest of
+     * {@code listFleet} -- a {@code workspace.list}/{@code tab.list} failure costs only the
+     * labels in the {@code panes} row (each renders {@code null}), never the {@code leads}/
+     * {@code members} arrays, which never needed that scan at all. fleetd #771: the same
+     * {@code workspace.list} failure costs {@code workspaceLabel} the same way.
+     */
+    @Test
+    void listStillReportsEveryOtherArrayWhenTheLabelScanFails() {
+        FakeHerdr h = new FakeHerdr().workspaceListFailsWith("unavailable");
+        FleetMcp.PaneSource panes = new FleetMcp.PaneSource(
+                () -> new PaneLocator(h).tabLabelsByTabId(),
+                () -> new PaneLocator(h).workspaceLabelsByWorkspaceId(),
+                Fleetd.deliverableTo(new MemberPresence(), Map::of, Map::of), _ -> false, _ -> false);
+
+        McpSchema.CallToolResult res = listFleetWithPanes(h, panes, true);
+
+        assertNotEquals(Boolean.TRUE, res.isError(), textOf(res));
+        String out = textOf(res);
+        assertTrue(out.contains("\"panes\":["), out);
+        assertTrue(out.contains("\"sessionId\":\"term_a\""), out);
+        assertTrue(out.contains("\"label\":null"), out);
+        assertTrue(out.contains("\"workspaceLabel\":null"),
+                "a workspace.list failure must not cost the leads/members arrays, only a null "
+                        + "workspaceLabel: " + out);
+        assertTrue(out.contains("\"leads\":[]"), "a label-scan failure must not cost the leads array: " + out);
+        assertTrue(out.contains("\"members\":[]"), "a label-scan failure must not cost the members array: " + out);
+    }
+
+    /**
+     * fleetd #756: a pane bound to a configured architect slot with no live member session must
+     * report {@code role: "architect"}, read from {@link CallerResolver#boundToArchitectSlot} —
+     * the same classifier {@link CallerResolver#sendableObserverTarget} refuses as a {@code SEND}
+     * target — rather than falling through to {@code "observer"}.
+     */
+    @Test
+    void listReportsArchitectForASlotBoundPaneWithNoLiveMember() {
+        FakeHerdr h = new FakeHerdr()
+                .withAgent("claude-arch", "term_unoccupied_architect", "w2:pArch", "w2:tArch");
+        FleetMcp.PaneSource panes = new FleetMcp.PaneSource(
+                Map::of, Map::of, _ -> false, "term_unoccupied_architect"::equals, _ -> false);
+
+        String out = textOf(listFleetWithPanes(h, panes, true));
+
+        assertTrue(out.contains("\"sessionId\":\"term_unoccupied_architect\""), out);
+        assertTrue(out.contains("\"role\":\"architect\""),
+                "a slot-bound pane with no live session must read \"architect\", not the generic "
+                        + "\"observer\" fallback: " + out);
+    }
+
+    /**
+     * Calls the canonical {@code listFleet} overload directly with an explicit {@code leads}/
+     * {@code collaborators} payload and {@code callerIsObserver}, mirroring exactly what the real
+     * {@code fleet_list} handler computes for an observer caller: {@code panesVisible} true,
+     * {@code leadsVisible}/{@code membersVisible}/{@code collaboratorsVisible} false.
+     */
+    private static McpSchema.CallToolResult listFleetAsObserver(FakeHerdr h, Map<String, String> leads,
+            Map<String, String> collaborators, FleetMcp.PaneSource panes) {
+        SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")));
+        return FleetMcp.listFleet(
+                workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
+                FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
+                FleetMcp.LoopHealthSource.none(), FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(),
+                FleetMcp.LeadSeatSource.none(), new LeadContextGauge(), FleetMcp.LeadConfigDirSource.none(),
+                leads, "", collaborators, false,
+                FleetMcp.CoordinationSource.none(), false, false, false, panes, true, true);
+    }
+
+    /**
+     * fleetd #758: an observer's {@code fleet_list} now carries a {@code panes} key, filtered to
+     * {@link CallerResolver#sendableObserverTarget} (so a lead's pane, a spawned member's pane, a
+     * collaborator's pane, and an unoccupied architect-slot pane are all absent) and every
+     * surviving row reduced to exactly {@code sessionId}, {@code label}, {@code status},
+     * {@code role}, {@code deliverable} — never {@code paneId}, {@code workspaceId},
+     * {@code workspaceLabel} (fleetd #771 — a space name is host shape, a stronger disclosure than
+     * a pane id, so it stays out of the reduced row too), {@code tabId}, {@code agentType}, or
+     * {@code cwd}.
+     */
+    @Test
+    void listFiltersAndReducesThePanesArrayForAnObserver() {
+        MemberRegistry members = new MemberRegistry(new FleetConfig.Fleet(Map.of(),
+                Map.of("lead-designer", new FleetConfig.Slot("sonnet")), Map.of(), Map.of(), null));
+        assertTrue(members.bind("architect:lead-designer", "term_architect_pane"));
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(null, false, null,
+                () -> Map.of("term_lead_pane", "fleet01-lead"), members,
+                t -> "term_member_pane".equals(t) ? MemberRole.DEV : null,
+                () -> Map.of("term_collab_pane", "ops"));
+
+        FakeHerdr h = new FakeHerdr()
+                .withAgent("claude-sendable", "term_sendable", "w2:pS", "w2:tS")
+                .withAgent("claude-lead", "term_lead_pane", "w2:pL", "w2:tL")
+                .withAgent("claude-member", "term_member_pane", "w2:pM", "w2:tM")
+                .withAgent("claude-collab", "term_collab_pane", "w2:pC", "w2:tC")
+                .withAgent("claude-arch", "term_architect_pane", "w2:pA", "w2:tA")
+                .withTab("w2", "w2:tS", "trinotes");
+
+        FleetMcp.PaneSource panes = new FleetMcp.PaneSource(
+                () -> new PaneLocator(h).tabLabelsByTabId(),
+                () -> new PaneLocator(h).workspaceLabelsByWorkspaceId(), _ -> true,
+                callers::boundToArchitectSlot, callers.sendableObserverTarget());
+
+        String out = textOf(listFleetAsObserver(h, callers.leads(),
+                Map.of("term_collab_pane", "ops"), panes));
+
+        assertTrue(out.contains("\"panes\":["), out);
+        assertTrue(out.contains("\"sessionId\":\"term_sendable\""),
+                "an ordinary unclassified pane must still be sendable and visible: " + out);
+        assertFalse(out.contains("term_lead_pane"), "a lead's pane must not be enumerated: " + out);
+        assertFalse(out.contains("term_member_pane"), "a spawned member's pane must not be enumerated: " + out);
+        assertFalse(out.contains("term_collab_pane"), "a collaborator's pane must not be enumerated: " + out);
+        assertFalse(out.contains("term_architect_pane"),
+                "an unoccupied architect-slot pane must not be enumerated: " + out);
+        assertFalse(out.contains("\"paneId\""), "an observer's row must never carry paneId: " + out);
+        assertFalse(out.contains("\"workspaceId\""), "an observer's row must never carry workspaceId: " + out);
+        assertFalse(out.contains("\"workspaceLabel\""),
+                "an observer's row must never carry workspaceLabel: " + out);
+        assertFalse(out.contains("\"tabId\""), "an observer's row must never carry tabId: " + out);
+        assertFalse(out.contains("\"agentType\""), "an observer's row must never carry agentType: " + out);
+        assertFalse(out.contains("\"cwd\""), "an observer's row must never carry cwd: " + out);
+    }
+
+    /**
+     * Control for the test above: a primary's {@code panes} row is unchanged by fleetd #758 —
+     * {@code callerIsObserver} false keeps every field, including {@code paneId} and a spawned
+     * member's {@code cwd}.
+     */
+    @Test
+    void listKeepsTheFullPaneRowForAPrimaryIncludingCwdAndPaneId() {
+        FakeHerdr h = new FakeHerdr();
+        SessionManager sessions = new SessionManager(workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")));
+        MemberSession spawned = sessions.acquire("ltms-local", "/worktree/member-1", null, null);
+        h.withAgent("claude-member", spawned.terminalId(), "w9:pMember", "w9:tMember");
+        FleetMcp.PaneSource panes = new FleetMcp.PaneSource(Map::of, Map::of, _ -> true, _ -> false, _ -> false);
+
+        McpSchema.CallToolResult res = FleetMcp.listFleet(
+                workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
+                FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
+                FleetMcp.LoopHealthSource.none(), FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(),
+                FleetMcp.LeadSeatSource.none(), new LeadContextGauge(), FleetMcp.LeadConfigDirSource.none(),
+                Map.of(), "", Map.of(), false,
+                FleetMcp.CoordinationSource.none(), true, true, true, panes, true, false);
+
+        String out = textOf(res);
+        assertTrue(out.contains("\"sessionId\":\"" + spawned.terminalId() + "\""), out);
+        assertTrue(out.contains("\"paneId\":\"w9:pMember\""),
+                "a primary must still see the fleet_stop handle: " + out);
+        assertTrue(out.contains("\"cwd\":\"/worktree/member-1\""),
+                "a primary must still see a spawned member's worktree path: " + out);
+        assertTrue(out.contains("\"role\":\"dev\""), out);
+    }
+
     /**
      * fleetd #421: {@code mailbox.pending} counts only broker-ready messages, so a blocked lead's
      * normal, healthy state is {@code "pending": 0} next to a non-empty {@code held[]} — which
@@ -900,7 +1379,7 @@ class FleetMcpTest {
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true);
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true, true, true);
 
         String out = textOf(res);
         assertTrue(out.contains("\"pending\":0"), out);
@@ -929,7 +1408,7 @@ class FleetMcpTest {
                 workerService(h, "http://gx00.gw:8000", Set.of("gx00.gw")), sessions, null,
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
-                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true);
+                Map.of(), "", new FleetMcp.CoordinationSource(channel, List.of()), true, true, true);
 
         String out = textOf(res);
         assertTrue(out.contains("\"heldDurable\":false"),
@@ -998,7 +1477,7 @@ class FleetMcpTest {
                 FleetMcp.CapacitySource.none(), new FleetMcp.HealthCoverageSource(() -> "off"),
                 FleetMcp.QuarantineSource.none(), FleetMcp.OutageSource.none(), FleetMcp.LeadSeatSource.none(),
                 Map.of(), "",
-                new FleetMcp.CoordinationSource(channel, List.of("fleet01-lead", "fleet02-lead", "fleet03-lead")), true);
+                new FleetMcp.CoordinationSource(channel, List.of("fleet01-lead", "fleet02-lead", "fleet03-lead")), true, true, true);
 
         String out = textOf(res);
         assertTrue(out.contains("\"coordId\":\"fleet01-lead\",\"status\":\"exists\",\"pending\":2,\"consumers\":1"), out);
@@ -1691,8 +2170,8 @@ class FleetMcpTest {
         Principal architect = Principal.architect("lead-designer", "term_design", 400);
         MemberPresence presence = new MemberPresence();
 
-        FleetMcp.markSpawnedMemberPresent(worker, presence);
-        FleetMcp.markSpawnedMemberPresent(architect, presence);
+        FleetMcp.markTrackedCallerPresent(worker, presence);
+        FleetMcp.markTrackedCallerPresent(architect, presence);
 
         assertTrue(presence.isPresent("term_worker"));
         assertTrue(presence.isPresent("term_design"));
@@ -1703,25 +2182,40 @@ class FleetMcpTest {
         Principal lead = Principal.leader("opus", "term_lead", 100);
         MemberPresence presence = new MemberPresence();
 
-        FleetMcp.markSpawnedMemberPresent(lead, presence);
-        FleetMcp.markSpawnedMemberPresent(Principal.anonymous(), presence);
+        FleetMcp.markTrackedCallerPresent(lead, presence);
+        FleetMcp.markTrackedCallerPresent(Principal.anonymous(), presence);
 
         assertFalse(presence.isPresent("term_lead"));
     }
 
+    /**
+     * The item whose absence would be silent: an observer's own MCP contact must still mark
+     * presence, or a pane resolving to the unconfigured-pane floor would sit on the injector
+     * readiness gate forever once something addresses it.
+     */
+    @Test
+    void anObserverContactMarksPresence() {
+        Principal observer = Principal.observer("term_observer", 800);
+        MemberPresence presence = new MemberPresence();
+
+        FleetMcp.markTrackedCallerPresent(observer, presence);
+
+        assertTrue(presence.isPresent("term_observer"));
+    }
+
     @Test
     void statusReportsLiveAgentStatus() {
         FakeHerdr blocked = new FakeHerdr().agentStatus("blocked");
         AgentControl blockedAgents = new AgentControl(blocked);
         McpSchema.CallToolResult res = FleetMcp.status(
-                new MessageService(blockedAgents, new Injector(blockedAgents), rendezvous), "term_a");
+                new MessageService(blockedAgents, new Injector(blockedAgents), rendezvous), "term_a", null);
         assertNotEquals(Boolean.TRUE, res.isError());
         assertEquals("blocked", textOf(res));
     }
 
     /**
-     * CB-582: a lead polling {@code fleet_status} on its normal cadence — not {@code fleet_poll}
-     * — must also see a worker's open async {@code fleet_ask} question, since the reverse-rendezvous
+     * A lead polling {@code fleet_status} on its normal cadence — not {@code fleet_poll} — must
+     * also see a worker's open async {@code fleet_ask} question, since the reverse-rendezvous
      * window it opened with is far shorter than that cadence.
      */
     @Test
@@ -1744,7 +2238,7 @@ class FleetMcpTest {
         } while (asking.phase() != MessageService.Phase.ASKING && System.currentTimeMillis() < deadline);
         assertEquals(MessageService.Phase.ASKING, asking.phase());
 
-        McpSchema.CallToolResult res = FleetMcp.status(messages, T);
+        McpSchema.CallToolResult res = FleetMcp.status(messages, T, null);
         assertNotEquals(Boolean.TRUE, res.isError());
         String out = textOf(res);
         assertTrue(out.startsWith("idle"), "the live status must still lead the text: " + out);
@@ -1755,7 +2249,73 @@ class FleetMcpTest {
         // Clean up the still-open ask so the background thread does not linger past the test.
         String turnId = asking.turnId();
         CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
-                () -> messages.answer(turnId, "config.yaml", 5000));
+                () -> messages.answer(turnId, "config.yaml", 5000, null));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        deadline = System.currentTimeMillis() + 3000;
+        while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
+            Thread.sleep(5);
+        }
+        assertTrue(rendezvous.resolve(T, "done"));
+        answer.get(5, TimeUnit.SECONDS);
+    }
+
+    /**
+     * {@code fleet_status}'s pending-ask block (the question, its {@code turnId} and its ticket)
+     * is shown only to the caller whose owner key created the delegation. An unnamed primary is
+     * held to the same rule: its owner key is {@code null}, which here does not match the named
+     * worker that created the delegation, so it sees none of the pending-ask fields either — the
+     * same as any other non-creating caller.
+     */
+    @Test
+    void statusGatesThePendingAskFieldsByTheDelegationsCreatorOwner() throws Exception {
+        Principal creator = Principal.worker("term_creator", 1);
+        String ticket = messages.sendAsync(T, "task that asks", null, creator);
+        long deadline = System.currentTimeMillis() + 3000;
+        while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
+            Thread.sleep(5);
+        }
+        assertTrue(rendezvous.isWaiting(T), "sendAsync should have opened its rendezvous waiter");
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
+
+        MessageService.TaskView asking;
+        deadline = System.currentTimeMillis() + 3000;
+        do {
+            asking = messages.poll(ticket);
+            Thread.sleep(5);
+        } while (asking.phase() != MessageService.Phase.ASKING && System.currentTimeMillis() < deadline);
+        assertEquals(MessageService.Phase.ASKING, asking.phase());
+
+        String other = textOf(FleetMcp.status(messages, T, "worker:term_other"));
+        assertTrue(other.startsWith("idle"), "the base status must still be shown: " + other);
+        assertFalse(other.contains("which config file?"),
+                "a non-creating caller must not see the question text: " + other);
+        assertFalse(other.contains(asking.turnId()),
+                "a non-creating caller must not see the turnId: " + other);
+        assertFalse(other.contains(ticket),
+                "a non-creating caller must not see the ticket: " + other);
+
+        String creatorStatus = textOf(FleetMcp.status(messages, T, creator.ownerKey()));
+        assertTrue(creatorStatus.contains("which config file?"),
+                "the creator must see the question: " + creatorStatus);
+        assertTrue(creatorStatus.contains(asking.turnId()),
+                "the creator must see the turnId: " + creatorStatus);
+        assertTrue(creatorStatus.contains(ticket), "the creator must see the ticket: " + creatorStatus);
+
+        String unnamed = textOf(FleetMcp.status(messages, T, null));
+        assertTrue(unnamed.startsWith("idle"), "the base status must still be shown: " + unnamed);
+        assertFalse(unnamed.contains("which config file?"),
+                "an unnamed primary must not see a question on a delegation a named worker created: " + unnamed);
+        assertFalse(unnamed.contains(asking.turnId()),
+                "a non-creating unnamed primary must not see the turnId: " + unnamed);
+        assertFalse(unnamed.contains(ticket),
+                "a non-creating unnamed primary must not see the ticket: " + unnamed);
+
+        // Clean up the still-open ask so the background thread does not linger past the test.
+        String turnId = asking.turnId();
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(turnId, "config.yaml", 5000, creator.ownerKey()));
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         deadline = System.currentTimeMillis() + 3000;
         while (!rendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
@@ -1837,6 +2397,53 @@ class FleetMcpTest {
         assertTrue(out.contains("\"sessionId\":\"term_design\""), out);
     }
 
+    /**
+     * A collaborator reports its own role and name, never the {@code leader} key a lead gets —
+     * {@code role} already reads {@code "collaborator"}, so a {@code leader} key alongside it
+     * would be self-contradicting. Control: the same call shape fed a named lead must still carry
+     * {@code leader}, so this is not passing because the key stopped being emitted for everyone.
+     */
+    @Test
+    void whoamiReportsACollaboratorWithNoLeaderKeyButALeadStillGetsOne() {
+        FakeHerdr h = new FakeHerdr();
+        SessionManager sessions = sessionManager(h, "http://gx00.gw:8000", Set.of("gx00.gw"));
+
+        McpSchema.CallToolResult collabRes = FleetMcp.whoami(
+                Principal.collaborator("ops", "term_collab", 700), sessions);
+        assertNotEquals(Boolean.TRUE, collabRes.isError());
+        String collabOut = textOf(collabRes);
+        assertTrue(collabOut.contains("\"role\":\"collaborator\""), collabOut);
+        assertTrue(collabOut.contains("\"collaborator\":\"ops\""), collabOut);
+        assertTrue(collabOut.contains("\"sessionId\":\"term_collab\""), collabOut);
+        assertFalse(collabOut.contains("leader"), collabOut);
+
+        McpSchema.CallToolResult leadRes = FleetMcp.whoami(
+                Principal.leader("opus", "term_lead", 100), sessions);
+        String leadOut = textOf(leadRes);
+        assertTrue(leadOut.contains("\"leader\":\"opus\""), leadOut);
+    }
+
+    /**
+     * An observer reports its own role and pane, never a {@code leader} key. Without an explicit
+     * branch it would reach the lead branch by elimination and look right only because the
+     * {@code leader} key is guarded on a non-null name — this pins the branch rather than the
+     * accident.
+     */
+    @Test
+    void whoamiReportsAnObserverNotALead() {
+        FakeHerdr h = new FakeHerdr();
+        SessionManager sessions = sessionManager(h, "http://gx00.gw:8000", Set.of("gx00.gw"));
+
+        McpSchema.CallToolResult res = FleetMcp.whoami(
+                Principal.observer("term_observer", 900), sessions);
+
+        assertNotEquals(Boolean.TRUE, res.isError());
+        String out = textOf(res);
+        assertTrue(out.contains("\"role\":\"observer\""), out);
+        assertTrue(out.contains("\"sessionId\":\"term_observer\""), out);
+        assertFalse(out.contains("leader"), out);
+    }
+
     /**
      * CB-548: an architect SEND delegates as its own pane (recording the per-target delegation) but
      * must NEVER become the legacy singleton "primary" fallback — the per-target map does not cure

fleetd/src/test/java/dev/ltms/fleet/mcp/PrimaryRegistryTest.java

@@ -169,4 +169,101 @@ class PrimaryRegistryTest {
         assertTrue(reg.nudgeTargetFor("term_worker").isEmpty());
         assertTrue(reg.nudgeTargetFor(null).isEmpty());
     }
+
+    // ── fleetd #737 unit 3: a named lead's terminal is resolved live, not just recorded ─────────
+
+    /**
+     * The whole point of carrying a name: a lead that has been rolled keeps its name but gets a
+     * fresh terminal. {@code currentTerminalForName} stands in for the live lead-tab scan here —
+     * it reports the lead now sits on a different terminal than the one that was recorded — and
+     * {@code nudgeTargetFor} must follow the name to that current terminal, not the stale one.
+     */
+    @Test
+    void nudgeTargetForFollowsARolledLeadsNameToItsCurrentTerminal() {
+        var reg = new PrimaryRegistry(null, name -> "opus".equals(name) ? "term_opus_after_roll" : null);
+        reg.recordDelegation("term_worker", "term_opus_before_roll", "opus");
+
+        assertEquals("term_opus_after_roll", reg.nudgeTargetFor("term_worker").orElseThrow(),
+                "the name must be resolved to the lead's CURRENT terminal, not the one recorded "
+                        + "at delegation time");
+    }
+
+    /**
+     * The lookup cannot place every name — an architect/collaborator name (never a lead), or a lead
+     * whose tab the scan cannot currently see (just rolled, off-host, non-herdr). Either way the
+     * terminal actually recorded is still the right thing to try, exactly as before this unit.
+     */
+    @Test
+    void nudgeTargetForFallsBackToTheRecordedTerminalWhenTheNameCannotBePlaced() {
+        var reg = new PrimaryRegistry(null, name -> null); // nothing is ever currently recognised
+        reg.recordDelegation("term_worker", "term_lead_recorded", "opus");
+
+        assertEquals("term_lead_recorded", reg.nudgeTargetFor("term_worker").orElseThrow());
+    }
+
+    /** The 2-arg {@code recordDelegation} overload records no name, so resolution never applies. */
+    @Test
+    void recordDelegationWithNoNameIsNeverResolvedByLookup() {
+        var reg = new PrimaryRegistry(null, name -> {
+            throw new AssertionError("a delegation recorded with no name must never consult the lookup");
+        });
+        reg.recordDelegation("term_worker", "term_lead");
+
+        assertEquals("term_lead", reg.nudgeTargetFor("term_worker").orElseThrow());
+    }
+
+    /** As {@link #nudgeTargetForFollowsARolledLeadsNameToItsCurrentTerminal}, for the singleton. */
+    @Test
+    void currentPrimaryTerminalFollowsARolledLeadsNameToItsCurrentTerminal() {
+        var reg = new PrimaryRegistry(null, name -> "sol".equals(name) ? "term_sol_after_roll" : null);
+        reg.record("term_sol_before_roll", "sol");
+
+        assertEquals("term_sol_after_roll", reg.currentPrimaryTerminal().orElseThrow());
+        assertEquals("term_sol_before_roll", reg.primaryTerminal().orElseThrow(),
+                "primaryTerminal() stays the raw recorded value — currentPrimaryTerminal() is the "
+                        + "one that resolves live");
+    }
+
+    @Test
+    void currentPrimaryTerminalFallsBackWhenTheNameCannotBePlaced() {
+        var reg = new PrimaryRegistry(null, name -> null);
+        reg.record("term_sol", "sol");
+
+        assertEquals("term_sol", reg.currentPrimaryTerminal().orElseThrow());
+    }
+
+    @Test
+    void currentPrimaryTerminalWithNoNameRecordedIsTheRawTerminal() {
+        var reg = new PrimaryRegistry(null, name -> {
+            throw new AssertionError("no name was ever recorded, the lookup must not be consulted");
+        });
+        reg.record("term_x");
+
+        assertEquals("term_x", reg.currentPrimaryTerminal().orElseThrow());
+    }
+
+    @Test
+    void currentPrimaryTerminalIsEmptyWhenNothingWasEverLearned() {
+        var reg = new PrimaryRegistry(null, name -> "anything");
+        assertTrue(reg.currentPrimaryTerminal().isEmpty());
+    }
+
+    /** A pin never carries a name, so a pinned registry's singleton resolution is always a no-op. */
+    @Test
+    void currentPrimaryTerminalForAPinIsNeverResolvedByLookup() {
+        var reg = new PrimaryRegistry("term_pinned", name -> {
+            throw new AssertionError("a pin carries no name, the lookup must not be consulted");
+        });
+
+        assertEquals("term_pinned", reg.currentPrimaryTerminal().orElseThrow());
+    }
+
+    /** {@code nudgeTargetFor}'s fallback to the singleton is the resolved one, not the raw one. */
+    @Test
+    void nudgeTargetForWithNoDelegationFallsBackToTheResolvedSingleton() {
+        var reg = new PrimaryRegistry(null, name -> "opus".equals(name) ? "term_opus_after_roll" : null);
+        reg.record("term_opus_before_roll", "opus");
+
+        assertEquals("term_opus_after_roll", reg.nudgeTargetFor("term_never_seen").orElseThrow());
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/msg/LeadCoordLoopTest.java

@@ -40,7 +40,7 @@ class LeadCoordLoopTest {
     @Test
     void deliversAHeldMessageToTheLeadPaneAndAcksIt() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "the merge is blocked"));
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
 
         loop(channel, herdr, Map.of(LEAD_TERM, SELF)).tick();
 
@@ -57,7 +57,7 @@ class LeadCoordLoopTest {
     @Test
     void redeliveryOfAMessageAlreadyWrittenToThePaneIsAckedWithoutAnotherPaneWrite() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "recover me"));
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
         var loop = loop(channel, herdr, Map.of(LEAD_TERM, SELF));
 
         loop.tick();
@@ -71,7 +71,7 @@ class LeadCoordLoopTest {
     @Test
     void aRedeliveryIsAckedEvenWhileTheLeadIsMidTurn() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "recover me"));
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
         var loop = loop(channel, herdr, Map.of(LEAD_TERM, SELF));
 
         loop.tick();
@@ -91,7 +91,7 @@ class LeadCoordLoopTest {
     @Test
     void leavesTheMessageUnackedWhenTheLeadIsMidTurn() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "hello"));
-        var herdr = new FakeHerdr().agentStatus("working");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("working");
 
         loop(channel, herdr, Map.of(LEAD_TERM, SELF)).tick();
 
@@ -103,7 +103,7 @@ class LeadCoordLoopTest {
     @Test
     void leavesTheMessageUnackedWhenNoLeadPaneIsKnown() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "hello"));
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
 
         loop(channel, herdr, Map.of()).tick();
 
@@ -115,7 +115,8 @@ class LeadCoordLoopTest {
     @Test
     void leavesTheMessageUnackedWhenHerdrRefusesTheInjection() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "hello"));
-        var herdr = new FakeHerdr().agentStatus("idle").agentSendFailsWith("agent_not_found");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET)
+                .agentStatus("idle").agentSendFailsWith("agent_not_found");
 
         loop(channel, herdr, Map.of(LEAD_TERM, SELF)).tick();
 
@@ -126,7 +127,7 @@ class LeadCoordLoopTest {
     @Test
     void resolvesTheLeadByNameWhenSeveralAreKnown() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "hello"));
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
         // Two leads on this daemon; only one carries the coord-id the mailbox is owned as.
         var leads = new java.util.LinkedHashMap<String, String>();
         leads.put("term_other", "some-other-lead");
@@ -142,7 +143,7 @@ class LeadCoordLoopTest {
     @Test
     void holdsWhenSeveralLeadsAreKnownAndNoneCarriesTheCoordId() {
         var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "hello"));
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
         var leads = new java.util.LinkedHashMap<String, String>();
         leads.put("term_one", "lead-one");
         leads.put("term_two", "lead-two");
@@ -159,7 +160,7 @@ class LeadCoordLoopTest {
         var channel = new FakeLeadChannel(SELF)
                 .hold(new LeadMessage("m1", PEER, SELF, "first"))
                 .hold(new LeadMessage("m2", PEER, SELF, "second"));
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
         var loop = loop(channel, herdr, Map.of(LEAD_TERM, SELF));
 
         loop.tick();
@@ -175,10 +176,51 @@ class LeadCoordLoopTest {
     @Test
     void anEmptyMailboxNeverTouchesHerdr() {
         var channel = new FakeLeadChannel(SELF);
-        var herdr = new FakeHerdr().agentStatus("idle");
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET).agentStatus("idle");
 
         loop(channel, herdr, Map.of(LEAD_TERM, SELF)).tick();
 
         assertEquals(0, herdr.calls.size(), "an idle fleet must not poll a pane's status every tick");
     }
+
+    @Test
+    void aLeadWithUnsubmittedTextInItsPromptBoxKeepsTheMessageHeldAndUnacked() {
+        var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "the merge is blocked"));
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.DRAFTED_PROMPT_CARET).agentStatus("idle");
+
+        loop(channel, herdr, Map.of(LEAD_TERM, SELF)).tick();
+
+        assertEquals(0, prompts(herdr).size(),
+                "delivery pastes and submits, so it must not land on a half-typed line");
+        assertEquals(List.of(), channel.acked(), "an undelivered message stays on the broker");
+        assertFalse(channel.peek().isEmpty(), "and is still held");
+    }
+
+    @Test
+    void aMessageHeldForADraftIsDeliveredOnALaterTick() {
+        var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "the merge is blocked"));
+        var herdr = new FakeHerdr().detectionText(FakeHerdr.DRAFTED_PROMPT_CARET).agentStatus("idle");
+        var loop = loop(channel, herdr, Map.of(LEAD_TERM, SELF));
+
+        loop.tick();
+        assertEquals(0, prompts(herdr).size());
+
+        herdr.detectionText(FakeHerdr.IDLE_PROMPT_CARET);
+        loop.tick();
+
+        assertEquals(1, prompts(herdr).size(), "the held message lands once the box is empty");
+        assertEquals(List.of("m1"), channel.acked());
+    }
+
+    @Test
+    void anUnreadablePaneKeepsTheMessageHeld() {
+        var channel = new FakeLeadChannel(SELF).hold(new LeadMessage("m1", PEER, SELF, "the merge is blocked"));
+        var herdr = new FakeHerdr().detectionText("garbled ansi noise with no input box").agentStatus("idle");
+
+        loop(channel, herdr, Map.of(LEAD_TERM, SELF)).tick();
+
+        assertEquals(0, prompts(herdr).size(), "a pane whose box cannot be found may be holding a draft");
+        assertEquals(List.of(), channel.acked());
+    }
+
 }

fleetd/src/test/java/dev/ltms/fleet/msg/LeadHeartbeatLoopTest.java

@@ -7,6 +7,7 @@ import ch.qos.logback.core.read.ListAppender;
 import com.fasterxml.jackson.databind.JsonNode;
 import com.fasterxml.jackson.databind.ObjectMapper;
 import dev.ltms.fleet.herdr.AgentControl;
+import dev.ltms.fleet.herdr.FakeHerdr;
 import dev.ltms.fleet.herdr.AgentStatus;
 import dev.ltms.fleet.herdr.HerdrClient;
 import dev.ltms.fleet.lead.LeadContextGauge;
@@ -648,6 +649,42 @@ class LeadHeartbeatLoopTest {
                 "the notice text must appear exactly once across all three sends: " + herdr.sentTexts());
     }
 
+    // ── fleetd #737 unit 3: tick() nudges the lead's CURRENT terminal, not the learned one ───────
+
+    /**
+     * {@code tick()} reads {@code primaryRegistry.currentPrimaryTerminal()} both to check the lead's
+     * status and to send the nudge. Here the registry learned the lead's terminal under its name
+     * before a roll; {@code currentTerminalForName} stands in for the live lead-tab scan and reports
+     * the lead now sits on a different terminal. A correct tick must follow the name and nudge the
+     * new terminal — nudging the old one would mean the heartbeat lost the lead across its own roll.
+     */
+    @Test
+    void tickNudgesTheLeadsCurrentTerminalAfterARoll() {
+        var herdr = new FailableHerdrClient("term_lead_after_roll");
+        var now = new AtomicLong(NOW);
+        InMemoryReplyInbox inbox = new InMemoryReplyInbox();
+        inbox.own(WORKER);
+        inbox.publish(WORKER, "m1", "hello");
+        @SuppressWarnings("unchecked")
+        List<MemberSession>[] rosterBox = new List[]{List.of(new MemberSession("p1", WORKER, "prof",
+                MemberRole.DEV, "/cwd", null, 0, 0, 0, MemberSession.State.READY, null, null))};
+        AgentControl agents = new AgentControl(herdr);
+        PrimaryRegistry registry = new PrimaryRegistry(null,
+                name -> "opus".equals(name) ? "term_lead_after_roll" : null);
+        registry.record("term_lead_before_roll", "opus");
+        ReplyPushLoop pushLoop = new ReplyPushLoop(registry, agents, inbox, scheduler, 5, 100_000);
+        LeadHeartbeatLoop loop = new LeadHeartbeatLoop(registry, agents, inbox, () -> rosterBox[0], pushLoop,
+                scheduler, now::get, IDLE_AFTER_NANOS, 100_000L, 0);
+
+        loop.tick(); // opens the idle window
+        now.addAndGet(TimeUnit.SECONDS.toNanos(400));
+        loop.tick(); // past the quiet period, pending reply -> INJECT
+
+        assertEquals(List.of("term_lead_after_roll"), herdr.promptTargets(),
+                "the heartbeat must read and nudge the lead's CURRENT terminal, not the one learned "
+                        + "before the roll");
+    }
+
     /**
      * Fake herdr client for the four tests above: always reports {@code lead} as IDLE, records the
      * {@code text} of every {@code agent.prompt} call, and can be told to throw on the very next
@@ -657,7 +694,10 @@ class LeadHeartbeatLoopTest {
         private static final ObjectMapper MAPPER = new ObjectMapper();
         private final String lead;
         private final List<String> sentTexts = new ArrayList<>();
+        private final List<String> promptTargets = new ArrayList<>();
         private boolean throwOnNextSend = false;
+        /** What {@code agent.read} reports — the loop reads the lead's input box before it nudges. */
+        private String paneTail = FakeHerdr.IDLE_PROMPT_CARET;
 
         FailableHerdrClient(String lead) {
             this.lead = lead;
@@ -667,10 +707,19 @@ class LeadHeartbeatLoopTest {
             throwOnNextSend = true;
         }
 
+        void paneTail(String tail) {
+            this.paneTail = tail;
+        }
+
         List<String> sentTexts() {
             return List.copyOf(sentTexts);
         }
 
+        /** Every terminal an {@code agent.prompt} call named, in call order. */
+        List<String> promptTargets() {
+            return List.copyOf(promptTargets);
+        }
+
         @Override
         @SuppressWarnings("unchecked")
         public JsonNode call(String method, Object params) {
@@ -680,13 +729,18 @@ class LeadHeartbeatLoopTest {
                                 .put("terminal_id", lead)
                                 .put("agent_status", "idle"));
             }
+            if ("agent.read".equals(method)) {
+                return MAPPER.createObjectNode()
+                        .set("read", MAPPER.createObjectNode().put("text", paneTail));
+            }
             if ("agent.prompt".equals(method)) {
+                Map<String, Object> p = params instanceof Map ? (Map<String, Object>) params : Map.of();
                 if (throwOnNextSend) {
                     throwOnNextSend = false;
                     throw new RuntimeException("simulated transient herdr send failure");
                 }
-                Map<String, Object> p = params instanceof Map ? (Map<String, Object>) params : Map.of();
                 sentTexts.add(String.valueOf(p.get("text")));
+                promptTargets.add(String.valueOf(p.get("target")));
             }
             return MAPPER.createObjectNode();
         }
@@ -695,4 +749,49 @@ class LeadHeartbeatLoopTest {
         public void close() {
         }
     }
+
+    // ── the operator's own prompt box ────────────────────────────────────────────────────────────
+
+    @Test
+    void tickHoldsTheNudgeWhileTheLeadsPromptBoxHoldsUnsubmittedText() {
+        var herdr = new FailableHerdrClient(LEAD);
+        herdr.paneTail(FakeHerdr.DRAFTED_PROMPT_CARET);
+        var now = new AtomicLong(NOW);
+        @SuppressWarnings("unchecked")
+        List<MemberSession>[] rosterBox = new List[]{List.of()};
+        InMemoryReplyInbox inbox = new InMemoryReplyInbox();
+        LeadHeartbeatLoop loop = tickableLoop(herdr, now, rosterBox, inbox, 0, scheduler);
+
+        loop.tick(); // opens the idle window
+        now.addAndGet(TimeUnit.SECONDS.toNanos(400));
+
+        loop.tick(); // would INJECT, but the operator is mid-sentence
+        assertEquals(0, herdr.sentTexts().size(),
+                "a nudge pastes and submits, so it must not land on a half-typed line");
+
+        herdr.paneTail(FakeHerdr.IDLE_PROMPT_CARET);
+        loop.tick();
+        assertEquals(1, herdr.sentTexts().size(), "the held nudge lands once the box is empty");
+        assertTrue(herdr.sentTexts().get(0).contains("Your own context is nearly full"),
+                "and it still carries the notice the held tick did not spend: " + herdr.sentTexts().get(0));
+    }
+
+    @Test
+    void anUnreadablePaneHoldsTheHeartbeatNudge() {
+        var herdr = new FailableHerdrClient(LEAD);
+        herdr.paneTail("garbled ansi noise with no input box");
+        var now = new AtomicLong(NOW);
+        @SuppressWarnings("unchecked")
+        List<MemberSession>[] rosterBox = new List[]{List.of()};
+        InMemoryReplyInbox inbox = new InMemoryReplyInbox();
+        LeadHeartbeatLoop loop = tickableLoop(herdr, now, rosterBox, inbox, 0, scheduler);
+
+        loop.tick();
+        now.addAndGet(TimeUnit.SECONDS.toNanos(400));
+        loop.tick();
+
+        assertEquals(0, herdr.sentTexts().size(),
+                "a pane whose box cannot be found may be holding a draft");
+    }
+
 }

fleetd/src/test/java/dev/ltms/fleet/msg/MessageServicePollUsageTest.java

@@ -0,0 +1,280 @@
+package dev.ltms.fleet.msg;
+
+import org.junit.jupiter.api.Test;
+
+import java.io.IOException;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.regex.Matcher;
+import java.util.regex.Pattern;
+import java.util.stream.Stream;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * Pins that no file under {@code src/main/java} calls the fail-open
+ * {@link MessageService#poll(String)} overload. That overload skips the ownership check in
+ * {@code MessageService}'s {@code ownsTicket} entirely, so a caller of it can read any session's
+ * ticket. Every production caller must go through {@link MessageService#poll(String, String)}
+ * and pass a {@code callerOwner} explicitly, even when it is {@code null}.
+ *
+ * <p>This reads each file's own source text rather than reflecting on compiled bytecode, because
+ * the risk is a future one-word edit at a call site, not a missing overload.
+ *
+ * <p>The scan below finds a violation by its receiver, {@code messages.poll(}, rather than the
+ * bare method name, so it does not mistake {@link java.util.Queue#poll()} for a violation. That
+ * anchor only covers a {@code MessageService} reached through a variable or field named
+ * {@code messages}, so {@link #everyMessageServiceDeclarationIsNamedMessages} pins the naming
+ * convention the anchor depends on: a declaration under any other name would be invisible to the
+ * scan above, and must turn this second check red instead of passing silently.
+ */
+class MessageServicePollUsageTest {
+
+    private static final Path PRODUCTION_SOURCE = Path.of("src/main/java");
+
+    @Test
+    void noProductionFileCallsTheSingleArgumentPollOverload() throws IOException {
+        List<String> violations = new ArrayList<>();
+        List<String> twoArgSites = new ArrayList<>();
+        int filesScanned = scanForPollCalls(PRODUCTION_SOURCE, violations, twoArgSites);
+
+        // CONTROL: the scan actually walked files -- a wrong root would otherwise report "no
+        // violations found" having looked at nothing.
+        assertTrue(filesScanned > 0, "control failed: the scan under " + PRODUCTION_SOURCE
+                + " visited zero .java files -- the path is wrong, so the absence of violations "
+                + "below proves nothing");
+
+        assertTrue(violations.isEmpty(), "found a call to the fail-open MessageService.poll(String) "
+                + "overload, which skips the ownership check entirely -- pass a callerOwner "
+                + "explicitly (even if null) through poll(String, String) instead: " + violations);
+
+        // CONTROL: the arity parser actually finds the two genuine two-argument call sites (the
+        // MCP handler in FleetMcp and the REST handler in FleetApp). If this drops, the parser
+        // itself is broken, not the production code -- a broken parser (or a scan root that
+        // reaches no real source) must fail loudly here rather than pass vacuously above.
+        assertEquals(2, twoArgSites.size(), "control failed: expected exactly the two known "
+                + "two-argument messages.poll(...) call sites, found: " + twoArgSites);
+        assertTrue(twoArgSites.stream().anyMatch(s -> s.contains("FleetMcp.java")),
+                "control failed: did not find the FleetMcp.java messages.poll(ticket, callerOwner) "
+                        + "site among: " + twoArgSites);
+        assertTrue(twoArgSites.stream().anyMatch(s -> s.contains("FleetApp.java")),
+                "control failed: did not find the FleetApp.java messages.poll(...) site among: "
+                        + twoArgSites);
+    }
+
+    /**
+     * The {@code messages.poll(} anchor above only sees a {@code MessageService} reached through
+     * a variable, field, or parameter named {@code messages}. This asserts that every such
+     * declaration under {@code src/main/java} uses that name, so a differently named declaration
+     * -- invisible to the scan above -- fails loudly here instead of letting that scan pass on a
+     * call site it never looked at.
+     */
+    @Test
+    void everyMessageServiceDeclarationIsNamedMessages() throws IOException {
+        List<String> names = new ArrayList<>();
+        int filesScanned = scanForDeclarationNames(PRODUCTION_SOURCE, names);
+
+        // CONTROL: the scan actually walked files -- a wrong root would otherwise report "every
+        // declaration is named messages" having looked at nothing.
+        assertTrue(filesScanned > 0, "control failed: the scan under " + PRODUCTION_SOURCE
+                + " visited zero .java files -- the path is wrong, so the result below proves nothing");
+
+        // CONTROL: the declaration pattern actually finds real declarations. Zero means the
+        // pattern is broken, not that every MessageService variable, field, or parameter vanished.
+        assertTrue(names.size() > 0, "control failed: found zero MessageService declarations under "
+                + PRODUCTION_SOURCE + " -- the declaration pattern is broken, update it before "
+                + "trusting the naming check below");
+
+        List<String> other = names.stream().filter(n -> !n.equals("messages")).distinct().toList();
+        assertTrue(other.isEmpty(), "found a MessageService declaration not named \"messages\": "
+                + other + " -- the messages.poll( scan above only looks for that name, so a call "
+                + "through a differently named variable or field is invisible to it; either rename "
+                + "the declaration or widen that scan's anchor to cover it");
+    }
+
+    private static int scanForPollCalls(Path root, List<String> violations, List<String> twoArgSites)
+            throws IOException {
+        List<Path> files = javaFiles(root);
+        for (Path file : files) {
+            scanFileForPollCalls(file, violations, twoArgSites);
+        }
+        return files.size();
+    }
+
+    private static int scanForDeclarationNames(Path root, List<String> names) throws IOException {
+        List<Path> files = javaFiles(root);
+        Pattern declaration = Pattern.compile("MessageService\\s+([A-Za-z_][A-Za-z0-9_]*)");
+        for (Path file : files) {
+            if (file.getFileName().toString().equals("MessageService.java")) {
+                continue; // the type's own declaration, not a caller holding a reference to it
+            }
+            String stripped = stripComments(Files.readString(file));
+            Matcher m = declaration.matcher(stripped);
+            while (m.find()) {
+                int j = m.end();
+                while (j < stripped.length() && Character.isWhitespace(stripped.charAt(j))) j++;
+                if (j < stripped.length() && stripped.charAt(j) == '(') {
+                    continue; // a method named like the convention, e.g. "MessageService messages()"
+                }
+                names.add(m.group(1));
+            }
+        }
+        return files.size();
+    }
+
+    private static List<Path> javaFiles(Path root) throws IOException {
+        try (Stream<Path> paths = Files.walk(root)) {
+            return paths.filter(p -> p.toString().endsWith(".java")).toList();
+        }
+    }
+
+    /**
+     * Replaces {@code //} and {@code /* *}{@code /} comment text with nothing, leaving code,
+     * string/char literals and line breaks untouched -- so a comment that merely mentions
+     * {@code MessageService} in prose can never be read as a declaration.
+     */
+    private static String stripComments(String source) {
+        StringBuilder out = new StringBuilder(source.length());
+        boolean inString = false;
+        boolean inChar = false;
+        int i = 0;
+        while (i < source.length()) {
+            char c = source.charAt(i);
+            if (inString) {
+                out.append(c);
+                if (c == '\\' && i + 1 < source.length()) { out.append(source.charAt(i + 1)); i += 2; continue; }
+                if (c == '"') inString = false;
+                i++;
+                continue;
+            }
+            if (inChar) {
+                out.append(c);
+                if (c == '\\' && i + 1 < source.length()) { out.append(source.charAt(i + 1)); i += 2; continue; }
+                if (c == '\'') inChar = false;
+                i++;
+                continue;
+            }
+            if (c == '"') { inString = true; out.append(c); i++; continue; }
+            if (c == '\'') { inChar = true; out.append(c); i++; continue; }
+            if (c == '/' && i + 1 < source.length() && source.charAt(i + 1) == '/') {
+                while (i < source.length() && source.charAt(i) != '\n') i++;
+                continue; // leaves the newline itself for the next iteration to append
+            }
+            if (c == '/' && i + 1 < source.length() && source.charAt(i + 1) == '*') {
+                i += 2;
+                while (i < source.length() && !(source.charAt(i) == '*' && i + 1 < source.length()
+                        && source.charAt(i + 1) == '/')) {
+                    if (source.charAt(i) == '\n') out.append('\n');
+                    i++;
+                }
+                i += 2;
+                continue;
+            }
+            out.append(c);
+            i++;
+        }
+        return out.toString();
+    }
+
+    private static void scanFileForPollCalls(Path file, List<String> violations, List<String> twoArgSites)
+            throws IOException {
+        String source = Files.readString(file);
+        String needle = "messages.poll(";
+        int from = 0;
+        int idx;
+        while ((idx = source.indexOf(needle, from)) >= 0) {
+            int argsStart = idx + needle.length();
+            String args = extractBalancedArgs(source, argsStart, file, idx);
+            int closeParenIndex = argsStart + args.length();
+            from = closeParenIndex + 1;
+
+            if (args.isBlank()) {
+                continue; // MessageService has no zero-argument poll() -- nothing to classify
+            }
+            String site = file + ":" + lineOf(source, idx);
+            if (topLevelCommaCount(args) == 0) {
+                violations.add(site + " -- messages.poll(" + args.trim() + ")");
+            } else {
+                twoArgSites.add(site);
+            }
+        }
+    }
+
+    /**
+     * The text between {@code messages.poll(} and its matching close paren: balanced over nested
+     * calls, and never split by a paren or comma sitting inside a string or char literal.
+     */
+    private static String extractBalancedArgs(String source, int start, Path file, int callIndex) {
+        int depth = 1;
+        boolean inString = false;
+        boolean inChar = false;
+        int i = start;
+        while (i < source.length()) {
+            char c = source.charAt(i);
+            if (inString) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '"') inString = false;
+            } else if (inChar) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '\'') inChar = false;
+            } else if (c == '"') {
+                inString = true;
+            } else if (c == '\'') {
+                inChar = true;
+            } else if (c == '(') {
+                depth++;
+            } else if (c == ')') {
+                depth--;
+                if (depth == 0) return source.substring(start, i);
+            }
+            i++;
+        }
+        throw new IllegalStateException(
+                "unbalanced parentheses scanning " + file + ":" + lineOf(source, callIndex));
+    }
+
+    /**
+     * Commas at paren/bracket/brace depth zero, skipping string and char literals -- the argument
+     * separators a human reader would see, not every comma character in the text.
+     */
+    private static int topLevelCommaCount(String args) {
+        int depth = 0;
+        int commas = 0;
+        boolean inString = false;
+        boolean inChar = false;
+        int i = 0;
+        while (i < args.length()) {
+            char c = args.charAt(i);
+            if (inString) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '"') inString = false;
+            } else if (inChar) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '\'') inChar = false;
+            } else if (c == '"') {
+                inString = true;
+            } else if (c == '\'') {
+                inChar = true;
+            } else if (c == '(' || c == '[' || c == '{') {
+                depth++;
+            } else if (c == ')' || c == ']' || c == '}') {
+                depth--;
+            } else if (c == ',' && depth == 0) {
+                commas++;
+            }
+            i++;
+        }
+        return commas;
+    }
+
+    private static int lineOf(String source, int index) {
+        int line = 1;
+        for (int i = 0; i < index; i++) {
+            if (source.charAt(i) == '\n') line++;
+        }
+        return line;
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/msg/MessageServiceTest.java

@@ -4,6 +4,7 @@ import ch.qos.logback.classic.Level;
 import ch.qos.logback.classic.Logger;
 import ch.qos.logback.classic.spi.ILoggingEvent;
 import ch.qos.logback.core.read.ListAppender;
+import dev.ltms.fleet.auth.Principal;
 import dev.ltms.fleet.herdr.AgentControl;
 import dev.ltms.fleet.herdr.AgentStatus;
 import dev.ltms.fleet.herdr.FakeHerdr;
@@ -78,7 +79,7 @@ class MessageServiceTest {
     }
 
     private CompletableFuture<MessageService.Reply> sendAsync(String content, long timeoutMillis) {
-        return CompletableFuture.supplyAsync(() -> messages.send(T, content, timeoutMillis));
+        return CompletableFuture.supplyAsync(() -> messages.send(T, content, timeoutMillis, null));
     }
 
     private void awaitWaiting() throws InterruptedException {
@@ -188,7 +189,7 @@ class MessageServiceTest {
                 localRendezvous, new InMemoryReplyInbox());
         String brief = "Implement the requested change. ".repeat(20);
         CompletableFuture<MessageService.Reply> send =
-                CompletableFuture.supplyAsync(() -> localMessages.send(T, brief, 5000));
+                CompletableFuture.supplyAsync(() -> localMessages.send(T, brief, 5000, (String) null));
         long deadline = System.currentTimeMillis() + 2000;
         while (!localRendezvous.isWaiting(T) && System.currentTimeMillis() < deadline) {
             Thread.sleep(5);
@@ -341,7 +342,7 @@ class MessageServiceTest {
 
         // The primary answers via fleet_send(turnId); this blocks again for the worker's reply.
         CompletableFuture<MessageService.Reply> answer =
-                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000));
+                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000, null));
 
         // The worker's ask returns the answer — it resumes the same turn.
         MessageService.AskResult a = ask.get(5, TimeUnit.SECONDS);
@@ -377,7 +378,7 @@ class MessageServiceTest {
 
         // The primary answers that one turnId; both asks unblock with the same answer.
         CompletableFuture<MessageService.Reply> answer =
-                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000));
+                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000, null));
 
         MessageService.AskResult a1 = ask1.get(5, TimeUnit.SECONDS);
         MessageService.AskResult a2 = ask2.get(5, TimeUnit.SECONDS);
@@ -472,7 +473,7 @@ class MessageServiceTest {
                 "the duplicate's own timeout elapses first");
 
         CompletableFuture<MessageService.Reply> answered =
-                CompletableFuture.supplyAsync(() -> messages.answer(turnId, "fleetd.yaml", 500));
+                CompletableFuture.supplyAsync(() -> messages.answer(turnId, "fleetd.yaml", 500, null));
 
         MessageService.AskResult a = fresh.get(5, TimeUnit.SECONDS);
         assertEquals(MessageService.AskOutcome.ANSWERED, a.outcome(),
@@ -507,7 +508,7 @@ class MessageServiceTest {
 
         CompletableFuture<MessageService.Reply> lateAnswer = new CompletableFuture<>();
         messages.setAskTimeoutRaceHookForTest(() ->
-                lateAnswer.complete(messages.answer(turnId, "too late", 500)));
+                lateAnswer.complete(messages.answer(turnId, "too late", 500, null)));
         try {
             MessageService.AskResult a = ask.get(5, TimeUnit.SECONDS);
             assertEquals(MessageService.AskOutcome.TIMED_OUT, a.outcome());
@@ -539,7 +540,7 @@ class MessageServiceTest {
 
     @Test
     void answeringAnUnknownTurnIsStale() {
-        MessageService.Reply r = messages.answer(T + "#999", "too late", 500);
+        MessageService.Reply r = messages.answer(T + "#999", "too late", 500, null);
         assertEquals(MessageService.Outcome.STALE_TURN, r.outcome(),
                 "an answer to a turn that never existed (or already lapsed) is stale, not a hang");
     }
@@ -572,7 +573,7 @@ class MessageServiceTest {
 
         messages.setAnswerAskLapseRaceHookForTest(() -> rendezvous.answerAsk(turnId, "raced in first"));
         try {
-            MessageService.Reply r = messages.answer(turnId, "too late", 500);
+            MessageService.Reply r = messages.answer(turnId, "too late", 500, null);
             assertEquals(MessageService.Outcome.STALE_TURN, r.outcome(),
                     "an ask already answered by the race must be seen as lapsed, not double-delivered");
         } finally {
@@ -596,7 +597,7 @@ class MessageServiceTest {
     void sendTimesOutBeforeDeliveryIsQueuedNotWorking() {
         // Nothing ever delivers the message and nothing resolves the send, so the reply future
         // times out with delivery still incomplete — the message is still queued for the worker.
-        MessageService.Reply r = messages.send(T, "never delivered", 50);
+        MessageService.Reply r = messages.send(T, "never delivered", 50, null);
         assertEquals(MessageService.Outcome.TIMED_OUT_QUEUED, r.outcome(),
                 "an undelivered send that times out is still queued, not working");
         assertNull(r.text());
@@ -605,7 +606,7 @@ class MessageServiceTest {
     @Test
     void sendTimesOutAfterDeliveryIsStillWorking() throws Exception {
         CompletableFuture<MessageService.Reply> send =
-                CompletableFuture.supplyAsync(() -> messages.send(T, "do the task", 300));
+                CompletableFuture.supplyAsync(() -> messages.send(T, "do the task", 300, null));
         awaitWaiting();
         injector.onStatus(T, AgentStatus.IDLE);    // deliver — the delivered future now completes
         injector.onStatus(T, AgentStatus.WORKING); // worker starts but never replies
@@ -630,7 +631,7 @@ class MessageServiceTest {
     void sendTimeoutUsesCancellationDeliveredWhenPickupWinsTheRace() {
         messages.setTimeoutCancellationRaceHookForTest(() -> injector.onStatus(T, AgentStatus.IDLE));
         try {
-            MessageService.Reply reply = messages.send(T, "race delivery", 50);
+            MessageService.Reply reply = messages.send(T, "race delivery", 50, null);
 
             assertEquals(MessageService.Outcome.TIMED_OUT_WORKING, reply.outcome(),
                     "cancel reporting DELIVERED means the worker received the timed-out message");
@@ -652,7 +653,7 @@ class MessageServiceTest {
     void sendTimesOutWithAttemptedDeliveryReportsUnconfirmedNotQueued() throws Exception {
         herdr.agentSendFailsWith("send_failed");
         CompletableFuture<MessageService.Reply> send =
-                CompletableFuture.supplyAsync(() -> messages.send(T, "brief", 150));
+                CompletableFuture.supplyAsync(() -> messages.send(T, "brief", 150, null));
         awaitWaiting();
         injector.onStatus(T, AgentStatus.IDLE); // triggers the failing delivery attempt → ATTEMPTED
 
@@ -679,7 +680,7 @@ class MessageServiceTest {
 
         // The primary answers, unblocking the worker; but the worker never sends the follow-up
         // fleet_reply, so the answering send rides out its short window as still-working.
-        MessageService.Reply answer = messages.answer(q.turnId(), "config.yaml", 200);
+        MessageService.Reply answer = messages.answer(q.turnId(), "config.yaml", 200, null);
         assertEquals(MessageService.Outcome.TIMED_OUT_WORKING, answer.outcome(),
                 "an answered worker that never replies times out as still working");
 
@@ -715,7 +716,7 @@ class MessageServiceTest {
         assertEquals(MessageService.Outcome.QUESTION, q.outcome());
 
         CompletableFuture<MessageService.Reply> answer =
-                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000));
+                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000, null));
         ask.get(5, TimeUnit.SECONDS); // worker resumed with the answer
 
         awaitWaiting(); // the answering call has (re)opened its own forward waiter
@@ -723,7 +724,7 @@ class MessageServiceTest {
         MessageService.Reply done = answer.get(5, TimeUnit.SECONDS);
         assertEquals(MessageService.Outcome.REPLIED, done.outcome());
 
-        MessageService.Reply probe = messages.send(T, "probe after normal reply", 300);
+        MessageService.Reply probe = messages.send(T, "probe after normal reply", 300, null);
         assertNotEquals(MessageService.Outcome.BUSY, probe.outcome(),
                 "the session lock must be released after a normal REPLIED answer(), or this bounded "
                         + "follow-up send would come back BUSY instead of timing out on its own work");
@@ -756,13 +757,13 @@ class MessageServiceTest {
         // The primary answers, unblocking the worker; the worker never sends its follow-up
         // fleet_reply, so the answering call rides out its short window as still-working.
         CompletableFuture<MessageService.Reply> answer =
-                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 200));
+                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 200, null));
         MessageService.Reply answered = answer.get(5, TimeUnit.SECONDS);
         assertEquals(MessageService.Outcome.TIMED_OUT_WORKING, answered.outcome(),
                 "an answered worker that never replies times out as still working");
         ask.get(5, TimeUnit.SECONDS); // drain: the worker resumed with the answer
 
-        MessageService.Reply probe = messages.send(T, "probe after timed-out-working", 300);
+        MessageService.Reply probe = messages.send(T, "probe after timed-out-working", 300, null);
         assertNotEquals(MessageService.Outcome.BUSY, probe.outcome(),
                 "the session lock must be released after a TIMED_OUT_WORKING answer(), or this bounded "
                         + "follow-up send would come back BUSY instead of timing out on its own work");
@@ -791,7 +792,7 @@ class MessageServiceTest {
                 new java.util.concurrent.atomic.AtomicReference<>();
         Thread answerer = new Thread(() -> {
             try {
-                messages.answer(turnId, "config.yaml", 5000);
+                messages.answer(turnId, "config.yaml", 5000, null);
                 caught.set(new AssertionError("expected answer() to throw"));
             } catch (Throwable t) {
                 caught.set(t);
@@ -813,7 +814,7 @@ class MessageServiceTest {
 
         ask.get(5, TimeUnit.SECONDS); // drain: the worker resumed with the answer
 
-        MessageService.Reply probe = messages.send(T, "probe after exceptional failure", 300);
+        MessageService.Reply probe = messages.send(T, "probe after exceptional failure", 300, null);
         assertNotEquals(MessageService.Outcome.BUSY, probe.outcome(),
                 "the session lock must be released when answer()'s reply future fails exceptionally, "
                         + "or this bounded follow-up send would come back BUSY instead of timing out on its own work");
@@ -840,7 +841,7 @@ class MessageServiceTest {
                 new java.util.concurrent.atomic.AtomicReference<>();
         Thread answerer = new Thread(() -> {
             try {
-                messages.answer(turnId, "config.yaml", 5000);
+                messages.answer(turnId, "config.yaml", 5000, null);
                 caught.set(new AssertionError("expected answer() to throw"));
             } catch (Throwable t) {
                 caught.set(t);
@@ -859,17 +860,212 @@ class MessageServiceTest {
 
         ask.get(5, TimeUnit.SECONDS); // drain: the worker resumed with the answer
 
-        MessageService.Reply probe = messages.send(T, "probe after interruption", 300);
+        MessageService.Reply probe = messages.send(T, "probe after interruption", 300, null);
         assertNotEquals(MessageService.Outcome.BUSY, probe.outcome(),
                 "the session lock must be released when answer()'s wait is interrupted, or this bounded "
                         + "follow-up send would come back BUSY instead of timing out on its own work");
     }
 
+    /**
+     * Drive an async send on {@code T} to a resolved reply, polling as {@code owner} until the
+     * ticket reports {@link MessageService.Phase#DONE} (or the 2s deadline runs out). {@code owner}
+     * must be a terminal this ticket's creator check actually accepts, or this loops to the
+     * deadline and returns a non-{@code DONE} view.
+     */
+    private MessageService.TaskView driveAsyncTicketToDone(String ticket, String owner) throws InterruptedException {
+        MessageService.TaskView view = null;
+        long deadline = System.currentTimeMillis() + 2000;
+        while (view == null || view.phase() != MessageService.Phase.DONE) {
+            if (System.currentTimeMillis() >= deadline) break;
+            view = messages.poll(ticket, owner);
+            //noinspection BusyWait
+            Thread.sleep(5);
+        }
+        return view;
+    }
+
     @Test
     void pollReturnsNullForAnUnknownTicket() {
         assertNull(messages.poll("task-999999"), "a ticket that was never minted is unknown");
     }
 
+    // --- fleetd #719: a per-boot nonce keeps one instance's ticket ids out of another's space ---
+
+    /** A second, fully independent instance — its own agents/injector/rendezvous/inbox, not shared. */
+    private MessageService newIndependentInstance() {
+        FakeHerdr otherHerdr = new FakeHerdr().readText("BUILD GREEN: 391 files");
+        AgentControl otherAgents = new AgentControl(otherHerdr);
+        Injector otherInjector = new Injector(otherAgents);
+        return new MessageService(otherAgents, otherInjector, new Rendezvous(), new InMemoryReplyInbox());
+    }
+
+    @Test
+    void twoInstancesMintDisjointTicketIds() {
+        MessageService other = newIndependentInstance();
+        String ticketFromThis = messages.sendAsync(T, "task on first instance", null, null);
+        String ticketFromOther = other.sendAsync(T, "task on second instance", null, null);
+        assertNotEquals(ticketFromThis, ticketFromOther,
+                "each instance mints its own id space, so even a first ticket from each must differ");
+    }
+
+    @Test
+    void foreignInstanceTicketDoesNotResolve() {
+        MessageService other = newIndependentInstance();
+        String ticket = messages.sendAsync(T, "task on first instance", null, null);
+        // `other` must reach the same sequence number, or this test passes against an empty map
+        // instead of against a colliding id.
+        other.sendAsync(T, "task on second instance", null, null);
+
+        // control: the id resolves in the instance that minted it, so a null below cannot be
+        // explained by broken plumbing — only by the ticket being foreign to `other`.
+        assertNotNull(messages.poll(ticket), "the minting instance must still resolve its own ticket");
+        assertNull(other.poll(ticket), "a ticket minted by a different instance must not resolve here");
+    }
+
+    // --- ticket ownership -----------------------------------------------------------------------
+
+    @Test
+    void leadBIsRefusedFromLeadAsTicket() throws Exception {
+        Principal leadA = Principal.leader("opus", "term_a", 1);
+        Principal leadB = Principal.leader("sol", "term_b", 2);
+        String ticket = messages.sendAsync(T, "long task", null, leadA);
+        awaitWaiting();
+        injector.onStatus(T, AgentStatus.IDLE);    // deliver
+        injector.onStatus(T, AgentStatus.WORKING); // worker works
+        assertTrue(rendezvous.resolve(T, "secret async result"), "a reply resolves the async send");
+
+        MessageService.TaskView owner = driveAsyncTicketToDone(ticket, leadA.ownerKey());
+        assertNotNull(owner, "the creator must still be able to read its own ticket");
+        assertEquals(MessageService.Phase.DONE, owner.phase());
+
+        MessageService.TaskView refused = messages.poll(ticket, leadB.ownerKey());
+        assertNotNull(refused, "a different lead gets a refusal, not silence");
+        assertNotEquals(MessageService.Phase.DONE, refused.phase(),
+                "a different lead must never see the ticket as DONE");
+        assertNull(refused.reply(), "a refusal must never carry the reply text");
+        assertFalse(String.valueOf(refused).contains("secret async result"),
+                "the reply text must not appear anywhere in the refused view");
+    }
+
+    @Test
+    void unnamedPrimaryIsRefusedFromANamedLeadsTicket() throws Exception {
+        String ticket = messages.sendAsync(T, "long task", null,
+                Principal.leader("opus", "term_lead", 1));
+        awaitWaiting();
+        injector.onStatus(T, AgentStatus.IDLE);    // deliver
+        injector.onStatus(T, AgentStatus.WORKING); // worker works
+        assertTrue(rendezvous.resolve(T, "primary-visible result"), "a reply resolves the async send");
+
+        MessageService.TaskView refused = messages.poll(ticket, Principal.primary(1).ownerKey());
+        assertNotNull(refused, "a different owner gets a refusal, not silence");
+        assertEquals(MessageService.Phase.FAILED, refused.phase());
+        assertEquals("forbidden: this ticket was created by a different session", refused.detail());
+        assertNull(refused.reply(), "a refusal must never carry the reply text");
+        assertFalse(String.valueOf(refused).contains("primary-visible result"),
+                "the reply text must not appear anywhere in the refused view");
+    }
+
+    /**
+     * Positive control for {@link #unnamedPrimaryIsRefusedFromANamedLeadsTicket}: without this,
+     * that test would pass just as well if {@code poll} refused every caller.
+     */
+    @Test
+    void unnamedPrimaryReadsItsOwnTicket() throws Exception {
+        String ticket = messages.sendAsync(T, "long task", null, Principal.primary(1));
+        awaitWaiting();
+        injector.onStatus(T, AgentStatus.IDLE);    // deliver
+        injector.onStatus(T, AgentStatus.WORKING); // worker works
+        assertTrue(rendezvous.resolve(T, "primary-visible result"), "a reply resolves the async send");
+
+        MessageService.TaskView view = driveAsyncTicketToDone(ticket, Principal.primary(2).ownerKey());
+        assertNotNull(view, "an unnamed primary must be able to read a ticket another unnamed primary created");
+        assertEquals(MessageService.Phase.DONE, view.phase());
+        assertEquals("primary-visible result", view.reply());
+    }
+
+    @Test
+    void theOneArgPollOverloadBypassesOwnershipEntirely() throws Exception {
+        String ticket = messages.sendAsync(T, "long task", null,
+                Principal.leader("opus", "term_lead", 1));
+        awaitWaiting();
+        injector.onStatus(T, AgentStatus.IDLE);    // deliver
+        injector.onStatus(T, AgentStatus.WORKING); // worker works
+        assertTrue(rendezvous.resolve(T, "primary-visible result"), "a reply resolves the async send");
+
+        MessageService.TaskView view = null;
+        long deadline = System.currentTimeMillis() + 2000;
+        while (view == null || view.phase() != MessageService.Phase.DONE) {
+            if (System.currentTimeMillis() >= deadline) break;
+            view = messages.poll(ticket); // the one-arg, no-check overload -- no caller owner key at all
+            //noinspection BusyWait
+            Thread.sleep(5);
+        }
+        assertNotNull(view, "the internal bypass must read a ticket owned by a named lead");
+        assertEquals(MessageService.Phase.DONE, view.phase());
+        assertEquals("primary-visible result", view.reply());
+    }
+
+    @Test
+    void namedLeadCanPollItsTicketAfterItsTerminalChanges() throws Exception {
+        Principal oldLead = Principal.leader("opus", "term_OLD", 1);
+        Principal newLead = Principal.leader("opus", "term_NEW", 2);
+        assertNotEquals(oldLead.terminal(), newLead.terminal(), "the test requires different terminals");
+        String ticket = messages.sendAsync(T, "long task", null, oldLead);
+        awaitWaiting();
+        injector.onStatus(T, AgentStatus.IDLE);    // deliver
+        injector.onStatus(T, AgentStatus.WORKING); // worker works
+        assertTrue(rendezvous.resolve(T, "own result"), "a reply resolves the async send");
+
+        MessageService.TaskView view = driveAsyncTicketToDone(ticket, newLead.ownerKey());
+        assertNotNull(view, "the same named lead must read the ticket from its new terminal");
+        assertEquals(MessageService.Phase.DONE, view.phase());
+        assertEquals("own result", view.reply());
+    }
+
+    @Test
+    void anonymousOwnerKeyIsRefusedByTheTicketGateItself() {
+        Principal lead = Principal.leader("opus", "term_lead", 1);
+        String ticket = messages.sendAsync(T, "long task", null, lead);
+
+        MessageService.TaskView refused = messages.poll(ticket, Principal.anonymous().ownerKey());
+
+        assertNotNull(refused);
+        assertEquals(MessageService.Phase.FAILED, refused.phase());
+        assertEquals("forbidden: this ticket was created by a different session", refused.detail());
+    }
+
+    /**
+     * {@link MessageService#isTicketOwnedBy} is the primitive {@code Authz}'s {@code TASK_READ}
+     * grant is built on, exposed so a caller can be granted read access before it ever polls.
+     * Paired with a negative control: the same ticket, a different creator's owner key.
+     */
+    @Test
+    void isTicketOwnedByMatchesOnlyTheRealCreator() {
+        Principal observerA = Principal.observer("term_observer_a", 1);
+        Principal observerB = Principal.observer("term_observer_b", 2);
+        String ticket = messages.sendAsync(T, "long task", null, observerA);
+
+        assertTrue(messages.isTicketOwnedBy(ticket, observerA.ownerKey()),
+                "the observer that created the ticket must be reported as its owner");
+        assertFalse(messages.isTicketOwnedBy(ticket, observerB.ownerKey()),
+                "a different observer must not be reported as the owner");
+    }
+
+    @Test
+    void isTicketOwnedByIsFalseForAnUnknownTicket() {
+        assertFalse(messages.isTicketOwnedBy("no-such-ticket", Principal.observer("term_x", 1).ownerKey()));
+    }
+
+    @Test
+    void architectOwnershipUsesTerminalRatherThanSlot() {
+        Principal oldArchitect = Principal.architect("opus", "term_OLD", 1);
+        Principal newArchitect = Principal.architect("opus", "term_NEW", 2);
+        String ticket = messages.sendAsync(T, "long task", null, oldArchitect);
+
+        assertEquals(MessageService.Phase.PENDING, messages.poll(ticket, oldArchitect.ownerKey()).phase());
+        assertEquals(MessageService.Phase.FAILED, messages.poll(ticket, newArchitect.ownerKey()).phase());
+    }
+
     @Test
     void pollReportsACompletedTicket() throws Exception {
         String ticket = messages.sendAsync(T, "long task");
@@ -896,11 +1092,11 @@ class MessageServiceTest {
     @Test
     void concurrentSendToSameSessionWhileFirstHoldsItIsBusy() throws Exception {
         CompletableFuture<MessageService.Reply> first =
-                CompletableFuture.supplyAsync(() -> messages.send(T, "first", 5000));
+                CompletableFuture.supplyAsync(() -> messages.send(T, "first", 5000, null));
         awaitWaiting(); // the first send now holds the session lock, blocked on its reply
 
         // A second send to the SAME session cannot take the lock within its short window.
-        MessageService.Reply busy = messages.send(T, "second", 100);
+        MessageService.Reply busy = messages.send(T, "second", 100, null);
         assertEquals(MessageService.Outcome.BUSY, busy.outcome(),
                 "a second send while another holds the session is busy, not a hang");
         assertNull(busy.text());
@@ -930,13 +1126,13 @@ class MessageServiceTest {
         PrimaryRegistry reg = new PrimaryRegistry(null);
         // L accepts a delegation to W: the send wins the lock and queues delivery → L is recorded.
         CompletableFuture<MessageService.Reply> first = CompletableFuture.supplyAsync(
-                () -> messages.send(T, "first", 5000, () -> reg.recordDelegation(T, LEAD_L)));
+                () -> messages.send(T, "first", 5000, () -> reg.recordDelegation(T, LEAD_L), null));
         awaitWaiting();
         assertEquals(LEAD_L, reg.nudgeTargetFor(T).orElseThrow(),
                 "an accepted send owns the delegation");
 
         // A attempts W while L holds it → BUSY (lock never taken) → its hook never fires.
-        MessageService.Reply busy = messages.send(T, "second", 100, () -> reg.recordDelegation(T, LEAD_A));
+        MessageService.Reply busy = messages.send(T, "second", 100, () -> reg.recordDelegation(T, LEAD_A), null);
         assertEquals(MessageService.Outcome.BUSY, busy.outcome());
         assertEquals(LEAD_L, reg.nudgeTargetFor(T).orElseThrow(),
                 "a BUSY send must not steal the delegator ownership it never earned");
@@ -957,7 +1153,7 @@ class MessageServiceTest {
     void anAcceptedSendAfterThePriorOwnerFinishesBecomesTheNewOwner() throws Exception {
         PrimaryRegistry reg = new PrimaryRegistry(null);
         CompletableFuture<MessageService.Reply> first = CompletableFuture.supplyAsync(
-                () -> messages.send(T, "first", 5000, () -> reg.recordDelegation(T, LEAD_L)));
+                () -> messages.send(T, "first", 5000, () -> reg.recordDelegation(T, LEAD_L), null));
         awaitWaiting();
         injector.onStatus(T, AgentStatus.IDLE);
         injector.onStatus(T, AgentStatus.WORKING);
@@ -968,7 +1164,7 @@ class MessageServiceTest {
 
         // L finished; A's later accepted send takes the delegation over.
         CompletableFuture<MessageService.Reply> second = CompletableFuture.supplyAsync(
-                () -> messages.send(T, "second", 5000, () -> reg.recordDelegation(T, LEAD_A)));
+                () -> messages.send(T, "second", 5000, () -> reg.recordDelegation(T, LEAD_A), null));
         awaitWaiting();
         assertEquals(LEAD_A, reg.nudgeTargetFor(T).orElseThrow(),
                 "an accepted send after the owner finished becomes the new delegator");
@@ -989,7 +1185,7 @@ class MessageServiceTest {
     void answeringAnAskDoesNotRewriteDelegatorOwnership() throws Exception {
         PrimaryRegistry reg = new PrimaryRegistry(null);
         CompletableFuture<MessageService.Reply> send = CompletableFuture.supplyAsync(
-                () -> messages.send(T, "do X", 5000, () -> reg.recordDelegation(T, LEAD_L)));
+                () -> messages.send(T, "do X", 5000, () -> reg.recordDelegation(T, LEAD_L), null));
         awaitWaiting();
         assertEquals(LEAD_L, reg.nudgeTargetFor(T).orElseThrow(), "L owns the delegation");
 
@@ -1004,7 +1200,7 @@ class MessageServiceTest {
 
         // L answers the ask on the same turn; the answer path must not touch ownership.
         CompletableFuture<MessageService.Reply> answer =
-                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000));
+                CompletableFuture.supplyAsync(() -> messages.answer(q.turnId(), "config.yaml", 5000, null));
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         awaitWaiting(); // the answering send reopened its forward waiter
         assertEquals(LEAD_L, reg.nudgeTargetFor(T).orElseThrow(),
@@ -1024,7 +1220,7 @@ class MessageServiceTest {
     void aThrowingAcceptedHookLeavesNoStaleWaiterOrQueuedOrphan() {
         assertThrows(IllegalStateException.class,
                 () -> messages.send(T, "doomed", 500,
-                        () -> { throw new IllegalStateException("ownership hook failed"); }),
+                        () -> { throw new IllegalStateException("ownership hook failed"); }, null),
                 "a throwing ownership hook fails the send loudly");
 
         assertFalse(rendezvous.isWaiting(T), "the failed send must not leave a stale rendezvous waiter");
@@ -1152,7 +1348,7 @@ class MessageServiceTest {
     @Test
     void abandonFailsASendThatIsStillWaitingOnAReleasedSession() throws Exception {
         CompletableFuture<MessageService.Reply> send =
-                CompletableFuture.supplyAsync(() -> messages.send(T, "work", 30_000));
+                CompletableFuture.supplyAsync(() -> messages.send(T, "work", 30_000, null));
         awaitWaiting();
 
         assertTrue(messages.abandon(T, "session released"), "a live waiter is abandoned");
@@ -1172,7 +1368,7 @@ class MessageServiceTest {
     @Test
     void abandonDoesNotOverwriteAnAlreadyResolvedSend() throws Exception {
         CompletableFuture<MessageService.Reply> send =
-                CompletableFuture.supplyAsync(() -> messages.send(T, "work", 30_000));
+                CompletableFuture.supplyAsync(() -> messages.send(T, "work", 30_000, null));
         awaitWaiting();
         assertTrue(rendezvous.resolve(T, "the real answer"));
 
@@ -1303,7 +1499,7 @@ class MessageServiceTest {
         assertEquals(MessageService.Phase.ASKING, messages.poll(ticket).phase());
 
         CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
-                () -> messages.answer(asking.turnId(), "config.yaml", 5000));
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, null));
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         awaitWaiting();
         assertTrue(rendezvous.resolve(T, "done"));
@@ -1343,7 +1539,7 @@ class MessageServiceTest {
         // same turnId must see it as lapsed rather than resolving a question nobody is waiting on.
         assertEquals(MessageService.AskOutcome.TIMED_OUT, ask.get(5, TimeUnit.SECONDS).outcome());
         assertEquals(MessageService.Outcome.STALE_TURN,
-                messages.answer(asking.turnId(), "config.yaml", 200).outcome());
+                messages.answer(asking.turnId(), "config.yaml", 200, null).outcome());
     }
 
     @Test
@@ -1380,7 +1576,7 @@ class MessageServiceTest {
 
         // The primary answers, but its own bounded wait for the worker's resumed turn is short and
         // expires before the worker (still genuinely working) gets back to it.
-        MessageService.Reply answerReply = messages.answer(asking.turnId(), "config.yaml", 150);
+        MessageService.Reply answerReply = messages.answer(asking.turnId(), "config.yaml", 150, null);
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         assertEquals(MessageService.Outcome.TIMED_OUT_WORKING, answerReply.outcome(),
                 "the primary's own bounded wait gives up before the worker finishes resuming");
@@ -1409,7 +1605,7 @@ class MessageServiceTest {
                 CompletableFuture.supplyAsync(() -> messages.ask(T, "which config?", 5000));
         MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
 
-        MessageService.Reply answerReply = messages.answer(asking.turnId(), "config.yaml", 150);
+        MessageService.Reply answerReply = messages.answer(asking.turnId(), "config.yaml", 150, null);
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         assertEquals(MessageService.Outcome.TIMED_OUT_WORKING, answerReply.outcome());
 
@@ -1463,7 +1659,7 @@ class MessageServiceTest {
         messages.setFinishAsyncTaskRaceHookForTest(() -> messages.forgetTurnForTest(turnId));
 
         CompletableFuture<MessageService.Reply> answer =
-                CompletableFuture.supplyAsync(() -> messages.answer(turnId, "config.yaml", 5000));
+                CompletableFuture.supplyAsync(() -> messages.answer(turnId, "config.yaml", 5000, null));
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         awaitWaiting(); // answer() opened its own forward waiter for the resumed worker turn
 
@@ -1569,7 +1765,7 @@ class MessageServiceTest {
         String turnId = asking.turnId();
 
         CompletableFuture<MessageService.Reply> answer =
-                CompletableFuture.supplyAsync(() -> messages.answer(turnId, "config.yaml", 5000));
+                CompletableFuture.supplyAsync(() -> messages.answer(turnId, "config.yaml", 5000, null));
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer(),
                 "the worker's own ask() call must have already unblocked with the primary's answer "
                         + "before we force the race below");
@@ -1616,7 +1812,7 @@ class MessageServiceTest {
         MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
         String turnId = asking.turnId();
 
-        MessageService.Reply answerReply = messages.answer(turnId, "config.yaml", 150);
+        MessageService.Reply answerReply = messages.answer(turnId, "config.yaml", 150, null);
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         assertEquals(MessageService.Outcome.TIMED_OUT_WORKING, answerReply.outcome(),
                 "the primary's own bounded wait must give up first, leaving turnId stamped with no "
@@ -1682,7 +1878,7 @@ class MessageServiceTest {
 
         MessageService.TaskView asking = messages.poll(first);
         CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
-                () -> messages.answer(asking.turnId(), "config.yaml", 5000));
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, null));
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
         awaitWaiting();
         assertTrue(rendezvous.resolve(T, "done"));
@@ -1716,7 +1912,7 @@ class MessageServiceTest {
 
         // The primary answers it — answer() resumes the turn and blocks for what comes next.
         CompletableFuture<MessageService.Reply> answer1 = CompletableFuture.supplyAsync(
-                () -> messages.answer(asking1.turnId(), "a1", 5000));
+                () -> messages.answer(asking1.turnId(), "a1", 5000, null));
         assertEquals("a1", ask1.get(5, TimeUnit.SECONDS).answer());
 
         // Still in the SAME resumed turn — before replying — the worker asks again.
@@ -1737,7 +1933,7 @@ class MessageServiceTest {
 
         // The primary answers the second question; the worker finally sends its real fleet_reply.
         CompletableFuture<MessageService.Reply> answer2 = CompletableFuture.supplyAsync(
-                () -> messages.answer(turnId2, "a2", 5000));
+                () -> messages.answer(turnId2, "a2", 5000, null));
         assertEquals("a2", ask2.get(5, TimeUnit.SECONDS).answer());
         awaitWaiting();
         assertTrue(rendezvous.resolve(T, "done"));
@@ -1804,20 +2000,20 @@ class MessageServiceTest {
         }
     }
 
-    // --- CB-582: fleet_status pendingAsk() ------------------------------------------------------
+    // --- fleet_status pendingAsk() -----------------------------------------------------------
 
     @Test
     void pendingAskReturnsNullWhenNoQuestionIsOpen() throws Exception {
-        assertNull(messages.pendingAsk(T), "no async ticket at all -> no pending ask");
+        assertNull(messages.pendingAsk(T, null), "no async ticket at all -> no pending ask");
 
         String ticket = messages.sendAsync(T, "long task");
         awaitWaiting();
-        assertNull(messages.pendingAsk(T), "a plain pending delegation is not a question");
+        assertNull(messages.pendingAsk(T, null), "a plain pending delegation is not a question");
 
         injectDelivery();
         assertTrue(rendezvous.resolve(T, "done"));
         awaitTicketPhase(ticket, MessageService.Phase.DONE);
-        assertNull(messages.pendingAsk(T), "a finished ticket carries no open question either");
+        assertNull(messages.pendingAsk(T, null), "a finished ticket carries no open question either");
     }
 
     @Test
@@ -1830,21 +2026,352 @@ class MessageServiceTest {
                 CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
         MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
 
-        MessageService.PendingAsk pending = messages.pendingAsk(T);
+        MessageService.PendingAsk pending = messages.pendingAsk(T, null);
         assertNotNull(pending, "fleet_status should see the open question");
         assertEquals(ticket, pending.ticket());
         assertEquals("which config file?", pending.question());
         assertEquals(asking.turnId(), pending.turnId());
 
         CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
-                () -> messages.answer(asking.turnId(), "config.yaml", 5000));
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, null));
         assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
-        assertNull(messages.pendingAsk(T), "an answered question is no longer pending");
+        assertNull(messages.pendingAsk(T, null), "an answered question is no longer pending");
         awaitWaiting();
         assertTrue(rendezvous.resolve(T, "done"));
         assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
     }
 
+    /**
+     * A caller's owner key must match the key that created the delegation to see its pending
+     * question. The unnamed primary is held to the same rule as everyone else: its key is
+     * {@code null}, which here does not match the named worker that created this delegation, so
+     * it is refused too.
+     */
+    @Test
+    void pendingAskGatesTheQuestionByTheDelegationsCreatorOwner() throws Exception {
+        Principal creator = Principal.worker("term_creator", 1);
+        String ticket = messages.sendAsync(T, "task that asks", null, creator);
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        assertNull(messages.pendingAsk(T, "worker:term_other"),
+                "a caller whose key did not create the delegation must not see the question");
+
+        MessageService.PendingAsk own = messages.pendingAsk(T, creator.ownerKey());
+        assertNotNull(own, "the creating caller must see its own open question");
+        assertEquals("which config file?", own.question());
+
+        assertNull(messages.pendingAsk(T, Principal.primary(1).ownerKey()),
+                "an unnamed primary must not see a question on a delegation a named worker created");
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, creator.ownerKey()));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    /**
+     * Positive control for {@link #pendingAskGatesTheQuestionByTheDelegationsCreatorOwner}:
+     * without this, that test's refusal would pass just as well if {@code pendingAsk} refused
+     * every caller. Here the delegation's creator is itself an unnamed primary (owner key
+     * {@code null}), so another unnamed primary's {@code null} key must still match it.
+     */
+    @Test
+    void unnamedPrimarySeesItsOwnDelegationsPendingQuestion() throws Exception {
+        String ticket = messages.sendAsync(T, "task that asks", null, Principal.primary(1));
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
+        awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        MessageService.PendingAsk unnamed = messages.pendingAsk(T, Principal.primary(2).ownerKey());
+        assertNotNull(unnamed, "an unnamed primary must see the question on a delegation another unnamed primary created");
+        assertEquals("which config file?", unnamed.question());
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(unnamed.turnId(), "config.yaml", 5000, null));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    /**
+     * {@code pendingAsk} has no public no-check overload the way {@link MessageService#poll}
+     * does, so this drives {@link MessageService#INTERNAL_NO_OWNER_CHECK} directly — the only way
+     * to exercise the bypass for this method.
+     */
+    @Test
+    void pendingAskInternalBypassSeesAnyDelegationsPendingQuestion() throws Exception {
+        Principal creator = Principal.worker("term_creator", 1);
+        String ticket = messages.sendAsync(T, "task that asks", null, creator);
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        MessageService.PendingAsk bypassed = messages.pendingAsk(T, MessageService.INTERNAL_NO_OWNER_CHECK);
+        assertNotNull(bypassed, "the internal bypass must see a question on a delegation a named worker created");
+        assertEquals("which config file?", bypassed.question());
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, creator.ownerKey()));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    /**
+     * A task created with no recorded owner (a short {@code sendAsync} overload) must not hand its
+     * open question to a caller with an owner key. Only the unnamed primary may still see it.
+     */
+    @Test
+    void pendingAskDeniesATerminalBearingCallerWhenTheTaskRecordsNoCreator() throws Exception {
+        String ticket = messages.sendAsync(T, "task that asks");
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        assertNull(messages.pendingAsk(T, "worker:term_someone"),
+                "a caller with an owner key must not see a question whose task records no creator");
+        assertNotNull(messages.pendingAsk(T, null),
+                "the unnamed primary must still see it even with no recorded creator");
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, null));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    // --- outstanding(): fleet_handover{open}'s list of open tickets and asks -------------------
+
+    @Test
+    void outstandingReportsOwnedPendingTicketsWithPhases() {
+        Principal lead = Principal.leader("opus", "term_lead", 1);
+        String ticket1 = messages.sendAsync(T, "task one", null, lead);
+        String ticket2 = messages.sendAsync(T, "task two", null, lead);
+
+        MessageService.Outstanding outstanding = messages.outstanding(lead.ownerKey());
+
+        assertEquals(2, outstanding.tickets().size());
+        assertTrue(outstanding.tickets().stream().anyMatch(t ->
+                        ticket1.equals(t.ticket()) && t.phase() == MessageService.Phase.PENDING && T.equals(t.target())),
+                "ticket1 must be reported PENDING: " + outstanding.tickets());
+        assertTrue(outstanding.tickets().stream().anyMatch(t ->
+                        ticket2.equals(t.ticket()) && t.phase() == MessageService.Phase.PENDING && T.equals(t.target())),
+                "ticket2 must be reported PENDING: " + outstanding.tickets());
+        assertTrue(outstanding.asks().isEmpty(), "neither ticket has an open question");
+    }
+
+    @Test
+    void outstandingReportsAnOpenAsksTurnId() throws Exception {
+        Principal lead = Principal.leader("opus", "term_lead", 1);
+        String ticket = messages.sendAsync(T, "task that asks", null, lead);
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        MessageService.Outstanding outstanding = messages.outstanding(lead.ownerKey());
+        assertEquals(1, outstanding.tickets().size());
+        assertEquals(MessageService.Phase.ASKING, outstanding.tickets().get(0).phase());
+        assertEquals(1, outstanding.asks().size());
+        MessageService.OutstandingAsk openAsk = outstanding.asks().get(0);
+        assertEquals(ticket, openAsk.ticket());
+        assertEquals(asking.turnId(), openAsk.turnId());
+        assertEquals(T, openAsk.workerSession());
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, lead.ownerKey()));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    /**
+     * Positive control: {@code leadA} must still see its own ticket and ask, so {@code leadB}
+     * seeing neither is the owner filter at work and not {@code outstanding} refusing everyone.
+     */
+    @Test
+    void outstandingDoesNotLeakAcrossNamedLeads() throws Exception {
+        Principal leadA = Principal.leader("opus", "term_a", 1);
+        Principal leadB = Principal.leader("sol", "term_b", 2);
+        String ticket = messages.sendAsync(T, "task that asks", null, leadA);
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        MessageService.Outstanding seenByA = messages.outstanding(leadA.ownerKey());
+        assertEquals(1, seenByA.tickets().size(), "lead A must see its own ticket");
+        assertEquals(1, seenByA.asks().size(), "lead A must see its own open ask");
+
+        MessageService.Outstanding seenByB = messages.outstanding(leadB.ownerKey());
+        assertTrue(seenByB.tickets().isEmpty(), "lead B must not see lead A's ticket");
+        assertTrue(seenByB.asks().isEmpty(), "lead B must not see lead A's open ask");
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, leadA.ownerKey()));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    @Test
+    void outstandingIsEmptyCollectionsNotNullForACallerWithNothing() {
+        MessageService.Outstanding outstanding =
+                messages.outstanding(Principal.leader("opus", "term_lead", 1).ownerKey());
+        assertNotNull(outstanding.tickets(), "a caller with no delegations still gets a list, not null");
+        assertNotNull(outstanding.asks(), "a caller with no delegations still gets a list, not null");
+        assertTrue(outstanding.tickets().isEmpty());
+        assertTrue(outstanding.asks().isEmpty());
+    }
+
+    /**
+     * A ticket is destroyed on a timer once it goes terminal ({@link #pruneTerminalTickets}'s TTL
+     * runs from completion), so it is the one case where carrying the id forward actually matters
+     * — a still-PENDING ticket is in no such danger, its worker is still running.
+     */
+    @Test
+    void outstandingReportsACompletedUncollectedTicketWithATerminalPhase() throws Exception {
+        Principal lead = Principal.leader("opus", "term_lead", 1);
+        String ticket = messages.sendAsync(T, "long task", null, lead);
+        awaitWaiting();
+        injectDelivery();
+        assertTrue(rendezvous.resolve(T, "async result"), "a reply resolves the async send");
+        awaitTicketPhase(ticket, MessageService.Phase.DONE);
+
+        MessageService.Outstanding outstanding = messages.outstanding(lead.ownerKey());
+        assertEquals(1, outstanding.tickets().size());
+        MessageService.OutstandingTicket done = outstanding.tickets().get(0);
+        assertEquals(ticket, done.ticket());
+        assertEquals(MessageService.Phase.DONE, done.phase());
+        assertEquals(T, done.target());
+    }
+
+    // --- fleetd #715: answer() is gated on the caller that owns the turn -----------------------
+
+    /**
+     * A blocking {@code fleet_send} from one caller opens the turn; a different caller's answer is
+     * refused with no side effect on the rendezvous or the worker's blocked {@code fleet_ask} —
+     * only the real owner can answer it.
+     */
+    @Test
+    void aDifferentCallersAnswerIsRefusedForABlockingSendDelegationButTheRealOwnerSucceeds() throws Exception {
+        CompletableFuture<MessageService.Reply> send = CompletableFuture.supplyAsync(
+                () -> messages.send(T, "do X", 5000, "term_owner"));
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config?", 5000));
+        MessageService.Reply q = send.get(5, TimeUnit.SECONDS);
+        assertEquals(MessageService.Outcome.QUESTION, q.outcome());
+        assertFalse(rendezvous.isWaiting(T), "the forward waiter closes once the question surfaces");
+
+        // The hijack: a different caller answers the SAME turnId.
+        MessageService.Reply hijacked = messages.answer(q.turnId(), "evil.yaml", 500, "term_attacker");
+        assertEquals(MessageService.Outcome.NOT_TURN_OWNER, hijacked.outcome(),
+                "a caller that did not open this turn must be refused, not served");
+        assertFalse(rendezvous.isWaiting(T), "a refused answer must not open a forward waiter");
+        assertFalse(ask.isDone(), "a refused answer must not resolve the worker's blocked fleet_ask");
+        assertEquals(T, rendezvous.askSession(q.turnId()), "a refused answer must leave the ask turn open");
+
+        // The control: the real owner answers the same turnId and the turn resumes normally.
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(q.turnId(), "config.yaml", 5000, "term_owner"));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    /**
+     * Same hijack and control as the blocking case, but the delegation is opened through the
+     * fire-and-poll path ({@code sendAsync}) — the owner comes from the ticket's recorded creator
+     * terminal, not a caller argument threaded through a live blocking call.
+     */
+    @Test
+    void aDifferentCallersAnswerIsRefusedForAnAsyncSendDelegationButTheRealOwnerSucceeds() throws Exception {
+        Principal owner = Principal.worker("term_owner", 1);
+        String ticket = messages.sendAsync(T, "task that asks", null, owner);
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        MessageService.Reply hijacked = messages.answer(asking.turnId(), "evil.yaml", 500,
+                "worker:term_attacker");
+        assertEquals(MessageService.Outcome.NOT_TURN_OWNER, hijacked.outcome(),
+                "a caller that did not create this delegation must be refused, not served");
+        assertFalse(ask.isDone(), "a refused answer must not resolve the worker's blocked fleet_ask");
+        assertEquals(MessageService.Phase.ASKING, messages.poll(ticket).phase(),
+                "a refused answer must not advance the async ticket's phase");
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, owner.ownerKey()));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+        assertEquals("done", awaitTicketPhase(ticket, MessageService.Phase.DONE).reply());
+    }
+
+    @Test
+    void namedLeadCanSeeAndAnswerAnAskAfterItsTerminalChangesWhileLeadBIsRefused() throws Exception {
+        Principal oldLead = Principal.leader("opus", "term_OLD", 1);
+        Principal newLead = Principal.leader("opus", "term_NEW", 2);
+        Principal leadB = Principal.leader("sol", "term_SOL", 3);
+        assertNotEquals(oldLead.terminal(), newLead.terminal(), "the test requires different terminals");
+        String ticket = messages.sendAsync(T, "task that asks", null, oldLead);
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config file?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        assertNotNull(messages.pendingAsk(T, newLead.ownerKey()),
+                "the same lead at its new terminal must see the pending ask");
+        assertNull(messages.pendingAsk(T, leadB.ownerKey()),
+                "another lead must not see the pending ask");
+        assertEquals(MessageService.Outcome.NOT_TURN_OWNER,
+                messages.answer(asking.turnId(), "evil.yaml", 500, leadB.ownerKey()).outcome());
+        assertFalse(ask.isDone(), "another lead must not resolve the worker's ask");
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000, newLead.ownerKey()));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+        assertEquals("done", awaitTicketPhase(ticket, MessageService.Phase.DONE).reply());
+    }
+
     // --- CB-588: async ticket terminal nudges ---------------------------------------------------
     //
     // MessageService.reply's rendezvous fast path is exactly what an async ticket always takes
@@ -1871,7 +2398,7 @@ class MessageServiceTest {
     private PushWiring wireWithPushLoop(int maxReminders, long backoffMs, java.util.function.LongSupplier nowNanos) {
         PrimaryRegistry registry = new PrimaryRegistry(null);
         registry.recordDelegation(T, LEAD);
-        FakeHerdr leadHerdr = new FakeHerdr();
+        FakeHerdr leadHerdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET);
         AgentControl leadAgents = new AgentControl(leadHerdr);
         var scheduler = java.util.concurrent.Executors.newSingleThreadScheduledExecutor();
         ReplyPushLoop pushLoop = new ReplyPushLoop(registry, leadAgents, inbox, scheduler, maxReminders, backoffMs);
@@ -1908,7 +2435,7 @@ class MessageServiceTest {
                                                       java.util.function.LongSupplier nowNanos) {
         PrimaryRegistry registry = new PrimaryRegistry(null);
         registry.recordDelegation(T, LEAD);
-        FakeHerdr leadHerdr = new FakeHerdr();
+        FakeHerdr leadHerdr = new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET);
         AgentControl leadAgents = new AgentControl(leadHerdr);
         ManualScheduler scheduler = new ManualScheduler();
         ReplyPushLoop pushLoop = new ReplyPushLoop(registry, leadAgents, inbox, scheduler, maxReminders, backoffMs);
@@ -2101,7 +2628,7 @@ class MessageServiceTest {
 
             // Clean up the still-open ask so the background thread does not linger past the test.
             CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
-                    () -> wiring.service().answer(asking.turnId(), "config.yaml", 5000));
+                    () -> wiring.service().answer(asking.turnId(), "config.yaml", 5000, null));
             assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
             awaitWaiting();
             assertTrue(rendezvous.resolve(T, "done"));
@@ -2127,7 +2654,7 @@ class MessageServiceTest {
                     .filter(c -> c.method().equals("agent.prompt")).count();
 
             CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
-                    () -> wiring.service().answer(asking.turnId(), "config.yaml", 5000));
+                    () -> wiring.service().answer(asking.turnId(), "config.yaml", 5000, null));
             assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
             awaitWaiting();
             java.util.concurrent.CountDownLatch terminalReached = new java.util.concurrent.CountDownLatch(1);
@@ -2167,7 +2694,8 @@ class MessageServiceTest {
         var scheduler = java.util.concurrent.Executors.newSingleThreadScheduledExecutor();
         // A backoff far longer than the test: the schedule is started but no tick ever fires, so
         // decide() is read directly and nothing here depends on timing.
-        ReplyPushLoop pushLoop = new ReplyPushLoop(registry, new AgentControl(new FakeHerdr()), inbox,
+        ReplyPushLoop pushLoop = new ReplyPushLoop(registry,
+                new AgentControl(new FakeHerdr().detectionText(FakeHerdr.IDLE_PROMPT_CARET)), inbox,
                 scheduler, 5, 60_000);
         MessageService service = new MessageService(agents, injector, rendezvous, inbox, pushLoop);
         try {
@@ -2520,7 +3048,7 @@ class MessageServiceTest {
     void hasQueuedDeliveryIsTrueAfterAnUndeliveredSendTimesOut() {
         // Nothing ever delivers the message (never goes IDLE/BLOCKED), so the send times out with
         // TIMED_OUT_QUEUED — same setup as sendTimesOutBeforeDeliveryIsQueuedNotWorking above.
-        MessageService.Reply r = messages.send(T, "never delivered", 50);
+        MessageService.Reply r = messages.send(T, "never delivered", 50, null);
         assertEquals(MessageService.Outcome.TIMED_OUT_QUEUED, r.outcome());
 
         assertTrue(messages.hasQueuedDelivery(T),
@@ -2532,7 +3060,7 @@ class MessageServiceTest {
 
     @Test
     void hasQueuedDeliveryClearsOnceTheTargetsNextDeliveryIsAccepted() throws Exception {
-        assertEquals(MessageService.Outcome.TIMED_OUT_QUEUED, messages.send(T, "first", 50).outcome());
+        assertEquals(MessageService.Outcome.TIMED_OUT_QUEUED, messages.send(T, "first", 50, null).outcome());
         assertTrue(messages.hasQueuedDelivery(T));
 
         // A fresh send accepts delivery (opens its own waiter) — the stale queued fact is cleared.
@@ -2549,7 +3077,7 @@ class MessageServiceTest {
 
     @Test
     void hasQueuedDeliveryClearsOnAbandon() {
-        assertEquals(MessageService.Outcome.TIMED_OUT_QUEUED, messages.send(T, "first", 50).outcome());
+        assertEquals(MessageService.Outcome.TIMED_OUT_QUEUED, messages.send(T, "first", 50, null).outcome());
         assertTrue(messages.hasQueuedDelivery(T));
 
         messages.abandon(T, "session released");

fleetd/src/test/java/dev/ltms/fleet/msg/RendezvousOpenUsageTest.java

@@ -0,0 +1,187 @@
+package dev.ltms.fleet.msg;
+
+import org.junit.jupiter.api.Test;
+
+import java.io.IOException;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.stream.Stream;
+
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * Pins that no file under {@code src/main/java} calls the fail-open
+ * {@link Rendezvous#open(String)} overload. That overload opens a forward waiter with no
+ * recorded {@link Rendezvous.Owner}, so the turn it opens can never be answered by anyone,
+ * not even the unnamed primary. Every production caller must go through
+ * {@link Rendezvous#open(String, Rendezvous.Owner)} and record an explicit owner.
+ *
+ * <p>This reads each file's own source text rather than reflecting on compiled bytecode, because
+ * the risk is a future one-word edit at a call site, not a missing overload.
+ *
+ * <p>The scan below finds a violation by its receiver, {@code rendezvous.open(}, classified by
+ * argument count. The one test method here points that exact scanner at a file known to hold
+ * many real one-argument calls before it ever looks at production, so a scanner that stops
+ * matching fails loudly on the known-positive case instead of leaving a clean production result
+ * looking like evidence it never produced.
+ */
+class RendezvousOpenUsageTest {
+
+    private static final Path PRODUCTION_SOURCE = Path.of("src/main/java");
+    private static final Path KNOWN_TEST_CALLER =
+            Path.of("src/test/java/dev/ltms/fleet/inject/CompletionResolverTest.java");
+
+    /**
+     * A pattern that cannot find a known one-argument {@code rendezvous.open(} call would also
+     * find none in production -- not because production is clean, but because the pattern does
+     * not match the text it is supposed to catch. That failure mode is exactly what let a
+     * {@code \b}-based regex read "no callers" under {@code git grep -E} when 81 real ones
+     * existed: {@code git grep} does not treat {@code \b} as a word boundary, so the pattern
+     * silently matched nothing anywhere, clean code and real calls alike. This test runs the
+     * known-positive check first, with the same scanning method the production check then
+     * depends on, so that mistake fails loudly here instead of reading as a clean result.
+     */
+    @Test
+    void noProductionFileCallsTheSingleArgumentOpenOverload() throws IOException {
+        List<String> knownCalls = new ArrayList<>();
+        int knownFilesScanned = scanForOneArgOpenCalls(KNOWN_TEST_CALLER, knownCalls);
+
+        // CONTROL: the scan actually walked files -- a wrong root would otherwise report "found
+        // nothing" having looked at nothing.
+        assertTrue(knownFilesScanned > 0, "control failed: the scan under " + KNOWN_TEST_CALLER
+                + " visited zero .java files -- the path is wrong, so neither result below proves "
+                + "anything");
+
+        // CONTROL: the scanner actually finds real one-argument rendezvous.open( calls when
+        // pointed at a file known to hold many. If this is not satisfied, the matching logic
+        // itself is broken, and the production result below is the scanner failing silently,
+        // not production code actually being clean.
+        assertTrue(knownCalls.size() >= 60, "control failed: the scanner found only "
+                + knownCalls.size() + " one-argument rendezvous.open( call(s) in " + KNOWN_TEST_CALLER
+                + ", which is known to hold many -- the matching logic itself is broken: " + knownCalls);
+
+        List<String> violations = new ArrayList<>();
+        int filesScanned = scanForOneArgOpenCalls(PRODUCTION_SOURCE, violations);
+
+        // CONTROL: the scan actually walked files -- a wrong root would otherwise report "no
+        // violations found" having looked at nothing.
+        assertTrue(filesScanned > 0, "control failed: the scan under " + PRODUCTION_SOURCE
+                + " visited zero .java files -- the path is wrong, so the absence of violations "
+                + "below proves nothing");
+
+        assertTrue(violations.isEmpty(), "found a call to the fail-open Rendezvous.open(String) "
+                + "overload, which opens a forward waiter with no recorded owner -- record an "
+                + "explicit Rendezvous.Owner through open(String, Owner) instead: " + violations);
+    }
+
+    private static int scanForOneArgOpenCalls(Path root, List<String> sites) throws IOException {
+        List<Path> files = javaFiles(root);
+        for (Path file : files) {
+            scanFileForOpenCalls(file, sites);
+        }
+        return files.size();
+    }
+
+    private static List<Path> javaFiles(Path root) throws IOException {
+        try (Stream<Path> paths = Files.walk(root)) {
+            return paths.filter(p -> p.toString().endsWith(".java")).toList();
+        }
+    }
+
+    private static void scanFileForOpenCalls(Path file, List<String> sites) throws IOException {
+        String source = Files.readString(file);
+        String needle = "rendezvous.open(";
+        int from = 0;
+        int idx;
+        while ((idx = source.indexOf(needle, from)) >= 0) {
+            int argsStart = idx + needle.length();
+            String args = extractBalancedArgs(source, argsStart, file, idx);
+            int closeParenIndex = argsStart + args.length();
+            from = closeParenIndex + 1;
+
+            if (args.isBlank()) {
+                continue; // Rendezvous has no zero-argument open() -- a javadoc "rendezvous.open()" mention, not a call
+            }
+            if (topLevelCommaCount(args) == 0) {
+                sites.add(file + ":" + lineOf(source, idx) + " -- rendezvous.open(" + args.trim() + ")");
+            }
+        }
+    }
+
+    /**
+     * The text between {@code rendezvous.open(} and its matching close paren: balanced over
+     * nested calls, and never split by a paren or comma sitting inside a string or char literal.
+     */
+    private static String extractBalancedArgs(String source, int start, Path file, int callIndex) {
+        int depth = 1;
+        boolean inString = false;
+        boolean inChar = false;
+        int i = start;
+        while (i < source.length()) {
+            char c = source.charAt(i);
+            if (inString) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '"') inString = false;
+            } else if (inChar) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '\'') inChar = false;
+            } else if (c == '"') {
+                inString = true;
+            } else if (c == '\'') {
+                inChar = true;
+            } else if (c == '(') {
+                depth++;
+            } else if (c == ')') {
+                depth--;
+                if (depth == 0) return source.substring(start, i);
+            }
+            i++;
+        }
+        throw new IllegalStateException(
+                "unbalanced parentheses scanning " + file + ":" + lineOf(source, callIndex));
+    }
+
+    /**
+     * Commas at paren/bracket/brace depth zero, skipping string and char literals -- the argument
+     * separators a human reader would see, not every comma character in the text.
+     */
+    private static int topLevelCommaCount(String args) {
+        int depth = 0;
+        int commas = 0;
+        boolean inString = false;
+        boolean inChar = false;
+        int i = 0;
+        while (i < args.length()) {
+            char c = args.charAt(i);
+            if (inString) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '"') inString = false;
+            } else if (inChar) {
+                if (c == '\\') { i += 2; continue; }
+                if (c == '\'') inChar = false;
+            } else if (c == '"') {
+                inString = true;
+            } else if (c == '\'') {
+                inChar = true;
+            } else if (c == '(' || c == '[' || c == '{') {
+                depth++;
+            } else if (c == ')' || c == ']' || c == '}') {
+                depth--;
+            } else if (c == ',' && depth == 0) {
+                commas++;
+            }
+            i++;
+        }
+        return commas;
+    }
+
+    private static int lineOf(String source, int index) {
+        int line = 1;
+        for (int i = 0; i < index; i++) {
+            if (source.charAt(i) == '\n') line++;
+        }
+        return line;
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/msg/RendezvousTest.java

@@ -132,4 +132,127 @@ class RendezvousTest {
         assertNull(rendezvous.askSession(t.turnId()), "a closed ask is forgotten");
         assertFalse(rendezvous.answerAsk(t.turnId(), "late"), "a closed ask can no longer be answered");
     }
+
+    // ── fleetd #715: turn ownership ────────────────────────────────────────────────────────────
+
+    @Test
+    void openWithNoOwnerRecordsNoOwnerAndOpenWithAnOwnerRecordsIt() {
+        rendezvous.open(W);
+        assertNull(rendezvous.ownerOf(W), "the no-owner overload records no owner at all");
+        rendezvous.close(W, rendezvous.currentWaiter(W));
+
+        rendezvous.open(W, Rendezvous.Owner.of("term_lead"));
+        assertEquals(Rendezvous.Owner.of("term_lead"), rendezvous.ownerOf(W));
+    }
+
+    @Test
+    void ownerOfIsNullWhenNoWaiterIsOpen() {
+        assertNull(rendezvous.ownerOf(W), "no waiter open means no owner to report");
+    }
+
+    @Test
+    void openAskStampsTheForwardWaitersOwnerOntoTheFreshTurnOnly() {
+        rendezvous.open(W, Rendezvous.Owner.of("term_lead"));
+
+        Rendezvous.AskTicket fresh = rendezvous.openAsk(W);
+        assertTrue(fresh.fresh());
+        assertEquals(Rendezvous.Owner.of("term_lead"), rendezvous.askOwner(fresh.turnId()),
+                "a freshly-opened ask copies the forward waiter's current owner");
+
+        Rendezvous.AskTicket coalesced = rendezvous.openAsk(W);
+        assertFalse(coalesced.fresh());
+        assertEquals(fresh.turnId(), coalesced.turnId());
+        assertEquals(Rendezvous.Owner.of("term_lead"), rendezvous.askOwner(coalesced.turnId()),
+                "a coalesced duplicate ask rides the fresh owner's turn, unchanged");
+    }
+
+    @Test
+    void aSecondAskAfterTheFirstClosesStampsWhateverOwnerIsOpenAtThatLaterMoment() {
+        rendezvous.open(W, Rendezvous.Owner.of("term_lead"));
+        Rendezvous.AskTicket first = rendezvous.openAsk(W);
+        rendezvous.closeAsk(first.turnId());
+
+        // The forward waiter is reopened under a different owner before the second ask — mirrors
+        // answer() reopening with the owner it already checked, which can differ turn to turn.
+        rendezvous.close(W, rendezvous.currentWaiter(W));
+        rendezvous.open(W, Rendezvous.Owner.of("term_other"));
+
+        Rendezvous.AskTicket second = rendezvous.openAsk(W);
+        assertTrue(second.fresh());
+        assertEquals(Rendezvous.Owner.of("term_other"), rendezvous.askOwner(second.turnId()),
+                "a freshly-opened ask always copies whatever owner is open right now, not a stale one");
+    }
+
+    @Test
+    void askOwnerIsNullForAnUnknownOrLapsedTurn() {
+        assertNull(rendezvous.askOwner("no-such#1"));
+        Rendezvous.AskTicket t = rendezvous.openAsk(W);
+        rendezvous.closeAsk(t.turnId());
+        assertNull(rendezvous.askOwner(t.turnId()), "a closed ask no longer reports an owner");
+    }
+
+    // ── fleetd #729: per-boot nonce guards turnId against cross-instance reuse ────────────────
+
+    @Test
+    void twoInstancesMintDisjointTurnIds() {
+        Rendezvous other = new Rendezvous();
+        Rendezvous.AskTicket fromThis = rendezvous.openAsk(W);
+        Rendezvous.AskTicket fromOther = other.openAsk(W);
+        assertNotEquals(fromThis.turnId(), fromOther.turnId(),
+                "each instance mints its own id space, so even a first ask from each must differ");
+    }
+
+    @Test
+    void foreignInstanceTurnIdDoesNotResolve() {
+        Rendezvous other = new Rendezvous();
+
+        // `other` must reach the same sequence number as `rendezvous` (two asks each, the first
+        // closed so the second mints fresh), or this test passes against an empty map instead of
+        // against a colliding id.
+        Rendezvous.AskTicket firstFromThis = rendezvous.openAsk(W);
+        rendezvous.closeAsk(firstFromThis.turnId());
+        Rendezvous.AskTicket secondFromThis = rendezvous.openAsk(W);
+
+        Rendezvous.AskTicket firstFromOther = other.openAsk(W);
+        other.closeAsk(firstFromOther.turnId());
+        other.openAsk(W);
+
+        // control: the id resolves in the instance that minted it, so a false below cannot be
+        // explained by broken plumbing — only by the turnId being foreign to `other`.
+        assertTrue(rendezvous.answerAsk(secondFromThis.turnId(), "answer from this instance"),
+                "the minting instance must still resolve its own turnId");
+        assertFalse(other.answerAsk(secondFromThis.turnId(), "answer from other instance"),
+                "a turnId minted by a different instance must not resolve here");
+    }
+
+    @Test
+    void openAskStillCoalescesDuplicatesAndStillMintsDistinctIdsPerAsk() {
+        Rendezvous.AskTicket t1 = rendezvous.openAsk(W);
+        Rendezvous.AskTicket t2 = rendezvous.openAsk(W);
+        assertEquals(t1.turnId(), t2.turnId(),
+                "a second openAsk while one is open still coalesces onto the same turn");
+        assertFalse(t2.fresh(), "the coalesced ask is still reported as not fresh");
+
+        rendezvous.closeAsk(t1.turnId());
+        Rendezvous.AskTicket t3 = rendezvous.openAsk(W);
+        assertNotEquals(t1.turnId(), t3.turnId(), "two asks from the same session still get different turnIds");
+    }
+
+    @Test
+    void ownerPermitsIsFailClosedOnARecordAndThreeStatesAreDistinct() {
+        assertFalse(Rendezvous.Owner.permits(null, null),
+                "no owner on record refuses even the unnamed primary");
+        assertFalse(Rendezvous.Owner.permits(null, "worker:term_a"),
+                "no owner on record refuses a caller with an owner key too");
+        assertTrue(Rendezvous.Owner.permits(Rendezvous.Owner.UNNAMED_PRIMARY, null),
+                "the recorded unnamed primary matches a caller with a null owner key");
+        assertFalse(Rendezvous.Owner.permits(Rendezvous.Owner.UNNAMED_PRIMARY, "worker:term_a"),
+                "the recorded unnamed primary does not match another owner key");
+        assertTrue(Rendezvous.Owner.permits(Rendezvous.Owner.of("worker:term_a"), "worker:term_a"),
+                "an owner matches the same key");
+        assertFalse(Rendezvous.Owner.permits(Rendezvous.Owner.of("worker:term_a"), "worker:term_b"),
+                "an owner refuses a different key");
+        assertFalse(Rendezvous.Owner.permits(Rendezvous.Owner.of("worker:term_a"), null),
+                "a named owner refuses the unnamed primary");
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/msg/ReplyPushLoopTest.java

@@ -3,6 +3,7 @@ package dev.ltms.fleet.msg;
 import com.fasterxml.jackson.databind.JsonNode;
 import com.fasterxml.jackson.databind.ObjectMapper;
 import dev.ltms.fleet.herdr.AgentControl;
+import dev.ltms.fleet.herdr.FakeHerdr;
 import dev.ltms.fleet.herdr.HerdrClient;
 import dev.ltms.fleet.herdr.HerdrException;
 import dev.ltms.fleet.mcp.PrimaryRegistry;
@@ -22,6 +23,7 @@ import java.util.concurrent.Executors;
 import java.util.concurrent.ScheduledExecutorService;
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicInteger;
+import java.util.function.BiPredicate;
 
 import static org.junit.jupiter.api.Assertions.*;
 
@@ -304,6 +306,40 @@ class ReplyPushLoopTest {
                         + "still live");
     }
 
+    // --- fleetd #737 unit 3: the fallback nudgeTargetFor returns must be probed too --------------
+
+    /**
+     * {@code resolveLiveLead} forgets a dead per-target binding and asks {@code PrimaryRegistry}
+     * again for a fallback. That fallback can be dead too — here PRIMARY, the pinned singleton, is
+     * affirmatively gone alongside DEAD_LEAD. A correct {@code resolveLiveLead} probes it exactly
+     * like the first lead and gives up for this tick rather than trust it unchecked.
+     *
+     * <p>This is checked through {@code onReplyQueued}/{@code isActive} rather than a send count:
+     * {@link ReplyPushLoop#onReplyQueued} only registers pending work and starts a schedule once
+     * {@code resolveLiveLead} returns a present value — a dead fallback that was trusted unprobed
+     * would already make this true, synchronously, with no tick or send needed to observe it. Pairs
+     * with {@link #aStaleLeadBindingFallsBackToTheLiveLeadInsteadOfNudgingADeadTerminal} as the
+     * positive control: same stale-DEAD_LEAD setup, but there the fallback (PRIMARY) is live and the
+     * nudge does fire — proving this test's "nothing happens" result comes from the fallback being
+     * dead, not from the assertion being unable to observe a nudge at all.
+     */
+    @Test
+    void aDoublyDeadFallbackIsNeverTrustedAndStartsNoSchedule() {
+        registry.recordDelegation(WORKER, DEAD_LEAD);
+
+        var rec = new AllDeadHerdrClient(Set.of(DEAD_LEAD, PRIMARY));
+        agents = new AgentControl(rec);
+        inbox.publish(WORKER, "m1", "hello");
+
+        var loop = loop(1, 50);
+        loop.onReplyQueued(WORKER);
+        assertFalse(loop.isActive(),
+                "both the per-target binding and the fallback are dead, so resolveLiveLead must "
+                        + "return empty and onReplyQueued must never register pending work or start "
+                        + "a schedule — an unprobed fallback would start one here");
+        assertEquals(0, rec.promptTargets().size(), "nobody live was found, so nothing was ever sent");
+    }
+
     // --- nudge format --------------------------------------------------------------------------
 
     @Test
@@ -448,6 +484,39 @@ class ReplyPushLoopTest {
         assertEquals(0, rec.sendCount(), "an already-collected ticket must never be nudged");
     }
 
+    /**
+     * A nudge must never instruct a target to run a call its own authorization would refuse — the
+     * push loop is a defensive check, not the source of truth for whether the call is allowed.
+     */
+    @Test
+    void aTicketIsNotNudgedWhenTheLeadMayNotPollIt() throws Exception {
+        var rec = recordingClient();
+        agents = new AgentControl(rec);
+        BiPredicate<String, String> neverAuthorized = (lead, ticket) -> false;
+
+        loop(1, 50, null, neverAuthorized).onTicketTerminal("task-1", WORKER, false);
+
+        Thread.sleep(300); // let the scheduled tick run
+        assertEquals(0, rec.sendCount(), "a ticket the lead may not poll must never be named in a nudge");
+    }
+
+    /**
+     * Positive control for the test above: the same setup, but a predicate that authorizes the
+     * call, still sends the nudge — proving the suppression above is the predicate's own doing,
+     * not some other change that silently stops every ticket nudge.
+     */
+    @Test
+    void aTicketIsNudgedWhenTheLeadMayPollIt() throws Exception {
+        var rec = recordingClient();
+        agents = new AgentControl(rec);
+        BiPredicate<String, String> alwaysAuthorized = (lead, ticket) -> true;
+
+        loop(1, 50, null, alwaysAuthorized).onTicketTerminal("task-1", WORKER, false);
+
+        assertTrue(rec.sendLatch.await(3, TimeUnit.SECONDS), "an authorized ticket must still be nudged");
+        assertEquals(1, rec.sendCount());
+    }
+
     @Test
     void ticketNudgesSendUpToCapThenStop() throws Exception {
         int cap = 2;
@@ -1045,6 +1114,76 @@ class ReplyPushLoopTest {
                 "hitting the ticket reminder cap must count as exhausted");
     }
 
+    // --- the operator's own prompt box ----------------------------------------------------------
+
+    @Test
+    void aLeadWithUnsubmittedTextInItsPromptBoxIsNotNudged() {
+        var herdr = new PaneTextHerdrClient(FakeHerdr.DRAFTED_PROMPT_CARET);
+        agents = new AgentControl(herdr);
+        inbox.publish(WORKER, "m1", "hello");
+        var loop = loop(5, 100_000);
+        loop.onReplyQueued(WORKER);
+
+        assertEquals(ReplyPushLoop.Action.WAIT_BUSY, loop.decide(PRIMARY, 0, 0),
+                "a nudge pastes and submits, so an idle lead mid-sentence must not be nudged");
+        assertEquals(0, herdr.promptCount(), "nothing reached the pane");
+        assertFalse(inbox.peek(WORKER).isEmpty(), "and the reply is still waiting to be collected");
+    }
+
+    @Test
+    void theSameLeadIsNudgedOnceItsPromptBoxIsEmpty() {
+        var herdr = new PaneTextHerdrClient(FakeHerdr.DRAFTED_PROMPT_CARET);
+        agents = new AgentControl(herdr);
+        inbox.publish(WORKER, "m1", "hello");
+        var loop = loop(5, 100_000);
+        loop.onReplyQueued(WORKER);
+
+        assertEquals(ReplyPushLoop.Action.WAIT_BUSY, loop.decide(PRIMARY, 0, 0));
+        herdr.paneText(FakeHerdr.IDLE_PROMPT_CARET);
+        assertEquals(ReplyPushLoop.Action.INJECT, loop.decide(PRIMARY, 0, 0),
+                "the box emptied, so the held nudge is due");
+    }
+
+    @Test
+    void anUnrecognisablePaneHoldsTheNudge() {
+        var herdr = new PaneTextHerdrClient("garbled ansi noise with no input box");
+        agents = new AgentControl(herdr);
+        inbox.publish(WORKER, "m1", "hello");
+        var loop = loop(5, 100_000);
+        loop.onReplyQueued(WORKER);
+
+        assertEquals(ReplyPushLoop.Action.WAIT_BUSY, loop.decide(PRIMARY, 0, 0),
+                "a pane whose box cannot be found may be holding a draft");
+        assertEquals(0, herdr.promptCount());
+    }
+
+    @Test
+    void aFailedPaneReadHoldsTheNudge() {
+        var herdr = new PaneTextHerdrClient(FakeHerdr.IDLE_PROMPT_CARET).failReads();
+        agents = new AgentControl(herdr);
+        inbox.publish(WORKER, "m1", "hello");
+        var loop = loop(5, 100_000);
+        loop.onReplyQueued(WORKER);
+
+        assertEquals(ReplyPushLoop.Action.WAIT_BUSY, loop.decide(PRIMARY, 0, 0),
+                "an unreadable box is treated as a draft, never as an empty one");
+        assertEquals(0, herdr.promptCount());
+    }
+
+    @Test
+    void aNudgeHeldForADraftIsSentOnALaterTick() throws Exception {
+        var herdr = new PaneTextHerdrClient(FakeHerdr.DRAFTED_PROMPT_CARET);
+        agents = new AgentControl(herdr);
+
+        loop(5, 50).onTicketTerminal("task-1", WORKER, false);
+
+        Thread.sleep(300);
+        assertEquals(0, herdr.promptCount(), "every tick holds while the operator is typing");
+        herdr.paneText(FakeHerdr.IDLE_PROMPT_CARET);
+        assertTrue(herdr.sendLatch.await(3, TimeUnit.SECONDS),
+                "the nudge lands on the first tick after the box empties");
+    }
+
     // --- helpers -------------------------------------------------------------------------------
 
     private ReplyPushLoop loop() {
@@ -1059,10 +1198,25 @@ class ReplyPushLoopTest {
         return new ReplyPushLoop(registry, agents, inbox, scheduler, maxReminders, backoffMs, metrics);
     }
 
+    private ReplyPushLoop loop(int maxReminders, long backoffMs, Metrics metrics,
+                                BiPredicate<String, String> ticketPollAuthorized) {
+        return new ReplyPushLoop(registry, agents, inbox, scheduler, maxReminders, backoffMs, metrics,
+                ticketPollAuthorized);
+    }
+
     private static AgentControl agentWithStatus(String status) {
         return new AgentControl(new FakeHerdrClient(status));
     }
 
+    /**
+     * The {@code agent.read} frame every fake here returns: a lead settled at an empty input box. The
+     * loop reads the box before it nudges, so a fake that answered nothing would read as a pane it
+     * cannot classify and hold every nudge.
+     */
+    private static JsonNode emptyPromptBoxRead() {
+        return MAPPER.createObjectNode().set("read", MAPPER.createObjectNode().put("text", FakeHerdr.IDLE_PROMPT_CARET));
+    }
+
     /** Non-recording (single-threaded) fake — safe for decide() tests. */
     private static final class FakeHerdrClient implements HerdrClient {
         private final String agentStatus;
@@ -1079,6 +1233,9 @@ class ReplyPushLoopTest {
                                 .put("terminal_id", PRIMARY)
                                 .put("agent_status", agentStatus));
             }
+            if ("agent.read".equals(method)) {
+                return emptyPromptBoxRead();
+            }
             return MAPPER.createObjectNode();
         }
 
@@ -1108,6 +1265,9 @@ class ReplyPushLoopTest {
                 calls.add(Map.entry(method, params));
                 sendLatch.countDown();
             }
+            if ("agent.read".equals(method)) {
+                return emptyPromptBoxRead();
+            }
             return MAPPER.createObjectNode();
         }
 
@@ -1145,6 +1305,9 @@ class ReplyPushLoopTest {
             if ("agent.prompt".equals(method)) {
                 sendLatch.countDown();
             }
+            if ("agent.read".equals(method)) {
+                return emptyPromptBoxRead();
+            }
             return MAPPER.createObjectNode();
         }
 
@@ -1182,6 +1345,9 @@ class ReplyPushLoopTest {
                 calls.add(Map.entry(method, params));
                 sendLatch.countDown();
             }
+            if ("agent.read".equals(method)) {
+                return emptyPromptBoxRead();
+            }
             return MAPPER.createObjectNode();
         }
 
@@ -1232,6 +1398,9 @@ class ReplyPushLoopTest {
                 promptTargets.add(String.valueOf(p.get("target")));
                 sendLatch.countDown();
             }
+            if ("agent.read".equals(method)) {
+                return emptyPromptBoxRead();
+            }
             return MAPPER.createObjectNode();
         }
 
@@ -1248,6 +1417,55 @@ class ReplyPushLoopTest {
         }
     }
 
+    /**
+     * Fake herdr client for fleetd #737 unit 3: every terminal named in {@code deadTargets} reports
+     * {@code agent_not_found} from {@code agent.get} — unlike {@link DeadLeadHerdrClient}, which can
+     * only make one terminal dead, this can make a per-target binding AND its fallback dead in the
+     * same test. {@code agent.prompt} is recorded unconditionally (no liveness check of its own),
+     * so a test can tell "resolveLiveLead probed and correctly found nobody live" (no prompt call)
+     * apart from "resolveLiveLead trusted a dead fallback and sent into it anyway" (a prompt call to
+     * a terminal this fake has already declared gone).
+     */
+    private static final class AllDeadHerdrClient implements HerdrClient {
+        private final Set<String> deadTargets;
+        private final List<String> promptTargets = Collections.synchronizedList(new ArrayList<>());
+
+        AllDeadHerdrClient(Set<String> deadTargets) {
+            this.deadTargets = deadTargets;
+        }
+
+        @Override
+        @SuppressWarnings("unchecked")
+        public JsonNode call(String method, Object params) {
+            Map<String, Object> p = params instanceof Map ? (Map<String, Object>) params : Map.of();
+            if ("agent.get".equals(method)) {
+                String target = String.valueOf(p.get("target"));
+                if (deadTargets.contains(target)) {
+                    throw new HerdrException("no such agent: " + target, "agent_not_found", null);
+                }
+                return MAPPER.createObjectNode()
+                        .set("agent", MAPPER.createObjectNode()
+                                .put("terminal_id", target)
+                                .put("agent_status", "idle"));
+            }
+            if ("agent.prompt".equals(method)) {
+                promptTargets.add(String.valueOf(p.get("target")));
+            }
+            if ("agent.read".equals(method)) {
+                return emptyPromptBoxRead();
+            }
+            return MAPPER.createObjectNode();
+        }
+
+        List<String> promptTargets() {
+            return List.copyOf(promptTargets);
+        }
+
+        @Override
+        public void close() {
+        }
+    }
+
     /**
      * Fake herdr client for fleetd #368 review: {@code flakyTarget}'s FIRST {@code agent.get} call
      * fails with a transient, non-{@code agent_not_found} {@code HerdrException} — a transport-level
@@ -1283,6 +1501,9 @@ class ReplyPushLoopTest {
                 promptTargets.add(String.valueOf(p.get("target")));
                 sendLatch.countDown();
             }
+            if ("agent.read".equals(method)) {
+                return emptyPromptBoxRead();
+            }
             return MAPPER.createObjectNode();
         }
 
@@ -1294,4 +1515,60 @@ class ReplyPushLoopTest {
         public void close() {
         }
     }
+
+    /**
+     * Thread-safe fake that always reports {@code idle} and serves a mutable pane tail, so a test can
+     * change what the lead's input box holds between ticks. Records every {@code agent.prompt}.
+     */
+    private static final class PaneTextHerdrClient implements HerdrClient {
+        private final List<Map.Entry<String, Object>> prompts =
+                Collections.synchronizedList(new ArrayList<>());
+        private volatile String paneText;
+        private volatile boolean failReads = false;
+        volatile CountDownLatch sendLatch = new CountDownLatch(1);
+
+        PaneTextHerdrClient(String paneText) {
+            this.paneText = paneText;
+        }
+
+        void paneText(String text) {
+            this.paneText = text;
+        }
+
+        PaneTextHerdrClient failReads() {
+            this.failReads = true;
+            return this;
+        }
+
+        int promptCount() {
+            return prompts.size();
+        }
+
+        @Override
+        public JsonNode call(String method, Object params) {
+            if ("agent.get".equals(method)) {
+                return MAPPER.createObjectNode()
+                        .set("agent", MAPPER.createObjectNode()
+                                .put("terminal_id", PRIMARY)
+                                .put("agent_status", "idle"));
+            }
+            if ("agent.read".equals(method)) {
+                if (failReads) {
+                    throw new HerdrException("herdr socket read timed out");
+                }
+                return MAPPER.createObjectNode()
+                        .set("read", MAPPER.createObjectNode().put("text", paneText));
+            }
+            if ("agent.prompt".equals(method)) {
+                prompts.add(Map.entry(method, params));
+                sendLatch.countDown();
+            }
+            return MAPPER.createObjectNode();
+        }
+
+        @Override
+        public void close() {
+        }
+    }
+
 }

fleetd/src/test/java/dev/ltms/fleet/rest/FleetAppAuthTest.java

@@ -1,8 +1,14 @@
 package dev.ltms.fleet.rest;
 
+import ch.qos.logback.classic.Level;
+import ch.qos.logback.classic.spi.ILoggingEvent;
+import com.fasterxml.jackson.databind.JsonNode;
+import com.fasterxml.jackson.databind.ObjectMapper;
 import dev.ltms.fleet.auth.CallerResolver;
 import dev.ltms.fleet.auth.Authz;
 import dev.ltms.fleet.auth.MemberRegistry;
+import dev.ltms.fleet.auth.Principal;
+import dev.ltms.fleet.auth.Role;
 import dev.ltms.fleet.config.FleetConfig;
 import dev.ltms.fleet.guard.SubscriptionGuard;
 import dev.ltms.fleet.herdr.AgentControl;
@@ -15,9 +21,11 @@ import dev.ltms.fleet.metrics.FleetMetrics;
 import dev.ltms.fleet.metrics.Metrics;
 import dev.ltms.fleet.msg.MessageService;
 import dev.ltms.fleet.msg.Rendezvous;
+import dev.ltms.fleet.peer.MemberRole;
 import dev.ltms.fleet.session.FakeWorktrees;
 import dev.ltms.fleet.session.SessionManager;
 import dev.ltms.fleet.member.ClaudeCodeLauncher;
+import dev.ltms.fleet.testing.CapturedLog;
 import io.javalin.Javalin;
 import org.junit.jupiter.api.AfterEach;
 import org.junit.jupiter.api.Test;
@@ -28,10 +36,15 @@ import java.net.http.HttpRequest;
 import java.net.http.HttpResponse;
 import java.nio.file.Files;
 import java.nio.file.Path;
+import java.util.ArrayList;
 import java.util.LinkedHashSet;
+import java.util.List;
 import java.util.Locale;
 import java.util.Map;
 import java.util.Set;
+import java.util.concurrent.CompletableFuture;
+import java.util.concurrent.TimeUnit;
+import java.util.function.Predicate;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 import java.util.stream.Collectors;
@@ -78,8 +91,13 @@ class FleetAppAuthTest {
         MessageService messages = new MessageService(agents, injector, new Rendezvous());
 
         ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> pid);
+        // term_a is the only herdr-owned pane this fixture's PID can resolve to (FakeHerdr's canned
+        // pane list), and this helper's own contract above says that pane is the worker -- so it
+        // must be recognised as a live spawned member here, the same way a real roster would,
+        // rather than falling through to the observer floor.
         CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, tokenMode, token,
-                Map::of, new MemberRegistry(null));
+                Map::of, new MemberRegistry(null), t -> "term_a".equals(t) ? MemberRole.DEV : null,
+                Map::of);
         metrics = FleetMetrics.create(sessions, new dev.ltms.fleet.msg.InMemoryReplyInbox());
 
         app = new FleetApp(herdr, workers, sessions, messages, sessions.asPresence(), null,
@@ -122,12 +140,503 @@ class FleetAppAuthTest {
         assertEquals(Authz.Action.DRAIN, FleetApp.routeAction("GET /sessions/{id}/replies"));
         assertEquals(Authz.Action.ASK, FleetApp.routeAction("POST /sessions/{id}/ask"));
         for (String route : Set.of("GET /sessions", "GET /agents", "GET /members", "GET /profiles",
-                "GET /member-credentials", "GET /sessions/{id}/status", "GET /tasks/{ticket}")) {
+                "GET /member-credentials")) {
             assertEquals(Authz.Action.READ, FleetApp.routeAction(route), route);
         }
+        for (String route : Set.of("GET /sessions/{id}/status", "GET /tasks/{ticket}")) {
+            assertEquals(Authz.Action.TASK_READ, FleetApp.routeAction(route), route);
+        }
         assertThrows(IllegalArgumentException.class, () -> FleetApp.routeAction("GET /healthz"));
     }
 
+    // --- GET /tasks/{ticket} must not be the no-check overload ----------------------------------
+
+    /**
+     * {@code GET /tasks/{ticket}} must resolve its caller the same way {@code allow(...)} does
+     * and thread that owner key into {@link MessageService#poll(String, String)}, not the
+     * no-check overload that ignores who is asking.
+     */
+    @Test
+    void theTaskStatusRouteActuallyThreadsTheCallersOwnerKeyIntoPoll() throws Exception {
+        String source = Files.readString(REST_SOURCE);
+
+        int start = source.indexOf("private void taskStatus(Context ctx) {");
+        assertTrue(start >= 0, "could not find taskStatus in " + REST_SOURCE
+                + " -- the scrape has stopped matching, fix the anchor before trusting this test");
+        int end = source.indexOf("private static void herdrError(Context ctx, HerdrException e) {", start);
+        assertTrue(end > start, "could not find the method declared after taskStatus to bound the scrape");
+        String handlerBlock = source.substring(start, end);
+
+        // CONTROL: the block we scraped really does call messages.poll(...) -- if this fails, the
+        // anchors above moved and the assertions below would otherwise pass on nothing.
+        assertTrue(handlerBlock.contains("messages.poll("),
+                "control failed: the scraped taskStatus block contains no messages.poll( call at all "
+                        + "-- the anchors have drifted, this test is not testing what it claims to");
+
+        assertTrue(handlerBlock.contains("caller.ownerKey()"),
+                "the taskStatus route must thread the resolved caller's owner key into messages.poll(...), "
+                        + "not the no-check overload -- block: " + handlerBlock);
+        assertTrue(handlerBlock.contains("ctx.attribute(CALLER)"),
+                "the taskStatus route must resolve its caller the same way allow(...) does, not via a "
+                        + "second, separate resolution path -- block: " + handlerBlock);
+    }
+
+    /**
+     * {@code GET /sessions/{id}/status} must resolve its caller the same way {@code allow(...)}
+     * does and thread that owner key into {@link MessageService#pendingAsk(String, String)}, not
+     * the no-check overload that ignores who is asking.
+     */
+    @Test
+    void theSessionStatusRouteActuallyThreadsTheCallersOwnerKeyIntoPendingAsk() throws Exception {
+        String source = Files.readString(REST_SOURCE);
+
+        int start = source.indexOf("private void sessionStatus(Context ctx) {");
+        assertTrue(start >= 0, "could not find sessionStatus in " + REST_SOURCE
+                + " -- the scrape has stopped matching, fix the anchor before trusting this test");
+        int end = source.indexOf("private void taskStatus(Context ctx) {", start);
+        assertTrue(end > start, "could not find the method declared after sessionStatus to bound the scrape");
+        String handlerBlock = source.substring(start, end);
+
+        // CONTROL: the block we scraped really does call messages.pendingAsk(...) -- if this fails,
+        // the anchors above moved and the assertions below would otherwise pass on nothing.
+        assertTrue(handlerBlock.contains("messages.pendingAsk("),
+                "control failed: the scraped sessionStatus block contains no messages.pendingAsk( "
+                        + "call at all -- the anchors have drifted, this test is not testing what it claims to");
+
+        assertTrue(handlerBlock.contains("caller.ownerKey()"),
+                "the sessionStatus route must thread the resolved caller's owner key into "
+                        + "messages.pendingAsk(...), not the no-check overload -- block: " + handlerBlock);
+        assertTrue(handlerBlock.contains("ctx.attribute(CALLER)"),
+                "the sessionStatus route must resolve its caller the same way allow(...) does, not via "
+                        + "a second, separate resolution path -- block: " + handlerBlock);
+    }
+
+    /**
+     * {@code GET /tasks/{ticket}} refuses a worker whose terminal did not create the ticket, and
+     * refuses an unnamed primary just the same: a named worker's ticket is not anyone else's to
+     * read, caller rank included. The ticket is minted directly on the shared
+     * {@link MessageService}, the same way {@code MessageServiceTest} drives
+     * {@link MessageService#poll(String, String)}, so this exercises only the REST poll route's
+     * own handling of the ownership already recorded on the ticket.
+     */
+    @Test
+    void restPollRefusesADifferentWorkerAndAnUnnamedPrimaryOnANamedWorkersTicket() throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        AgentControl agents = new AgentControl(herdr);
+        Injector injector = new Injector(agents);
+        MessageService messages = new MessageService(agents, injector, new Rendezvous());
+
+        Javalin creatorApp = startOnSharedService(messages, herdr, FakeHerdr.WORKER_PID); // -> term_a
+        Javalin otherWorkerApp = startOnSharedService(messages, herdr, 9001L);             // -> term_shell
+        Javalin primaryApp = startOnSharedService(messages, herdr, 999_999L);              // no pane -> primary
+        try {
+            String ticket = messages.sendAsync("term_a", "long task", null,
+                    Principal.worker("term_a", FakeHerdr.WORKER_PID));
+
+            HttpResponse<String> refused = send(otherWorkerApp.port(), "GET", "/tasks/" + ticket, null, null);
+            assertEquals(200, refused.statusCode());
+            assertTrue(refused.body().contains("forbidden"),
+                    "a different worker's terminal must be refused, not shown the ticket: " + refused.body());
+            assertFalse(refused.body().contains("\"reply\""),
+                    "a refusal must never carry reply text: " + refused.body());
+
+            HttpResponse<String> own = send(creatorApp.port(), "GET", "/tasks/" + ticket, null, null);
+            assertEquals(200, own.statusCode());
+            assertFalse(own.body().contains("forbidden"),
+                    "the creating worker must read its own ticket: " + own.body());
+
+            HttpResponse<String> primary = send(primaryApp.port(), "GET", "/tasks/" + ticket, null, null);
+            assertEquals(200, primary.statusCode());
+            assertTrue(primary.body().contains("forbidden"),
+                    "an unnamed primary must not read a ticket a named worker created: " + primary.body());
+            assertFalse(primary.body().contains("\"reply\""),
+                    "a refusal must never carry reply text: " + primary.body());
+        } finally {
+            creatorApp.stop();
+            otherWorkerApp.stop();
+            primaryApp.stop();
+        }
+    }
+
+    /**
+     * Positive control for
+     * {@link #restPollRefusesADifferentWorkerAndAnUnnamedPrimaryOnANamedWorkersTicket}: without
+     * this, that test's refusal would pass just as well if the route refused every caller. Here
+     * the ticket's creator is itself an unnamed primary, so another unnamed primary reading it
+     * over REST must still succeed.
+     */
+    @Test
+    void restPollAllowsAnUnnamedPrimaryItsOwnTicket() throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        AgentControl agents = new AgentControl(herdr);
+        Injector injector = new Injector(agents);
+        MessageService messages = new MessageService(agents, injector, new Rendezvous());
+
+        Javalin primaryApp = startOnSharedService(messages, herdr, 999_999L); // no pane -> primary
+        try {
+            String ticket = messages.sendAsync("term_a", "long task", null, Principal.primary(FakeHerdr.WORKER_PID));
+
+            HttpResponse<String> own = send(primaryApp.port(), "GET", "/tasks/" + ticket, null, null);
+            assertEquals(200, own.statusCode());
+            assertFalse(own.body().contains("forbidden"),
+                    "an unnamed primary must read a ticket another unnamed primary created: " + own.body());
+        } finally {
+            primaryApp.stop();
+        }
+    }
+
+    /**
+     * {@code POST /sessions/{id}/message} with {@code wait:false} must record the creating
+     * caller's own terminal on the ticket it returns, so that caller can still poll its own
+     * ticket over REST, while a different terminal is refused.
+     */
+    @Test
+    void restSendAsyncRecordsTheCreatingCallersTerminalSoItCanStillPollItsOwnTicket() throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        AgentControl agents = new AgentControl(herdr);
+        Injector injector = new Injector(agents);
+        MessageService messages = new MessageService(agents, injector, new Rendezvous());
+
+        Javalin leadApp = startOnSharedService(messages, herdr, FakeHerdr.WORKER_PID, Map.of("term_a", "lead-x"));
+        Javalin otherWorkerApp = startOnSharedService(messages, herdr, 9001L); // -> term_shell
+        try {
+            ObjectMapper mapper = new ObjectMapper();
+            HttpResponse<String> created = send(leadApp.port(), "POST", "/sessions/term_a/message",
+                    "{\"content\":\"long task\",\"wait\":false}", null);
+            assertEquals(202, created.statusCode(), created.body());
+            String ticket = mapper.readTree(created.body()).path("ticket").asText(null);
+            assertNotNull(ticket, "the accepted response carried no ticket: " + created.body());
+
+            HttpResponse<String> own = send(leadApp.port(), "GET", "/tasks/" + ticket, null, null);
+            assertEquals(200, own.statusCode());
+            assertFalse(own.body().contains("forbidden"),
+                    "the session that created the ticket over REST must be able to poll it: " + own.body());
+
+            HttpResponse<String> refused = send(otherWorkerApp.port(), "GET", "/tasks/" + ticket, null, null);
+            assertEquals(200, refused.statusCode());
+            assertTrue(refused.body().contains("forbidden: this ticket was created by a different session"),
+                    "a different terminal must still be refused with the ownership detail, not some "
+                            + "other rejection: " + refused.body());
+        } finally {
+            leadApp.stop();
+            otherWorkerApp.stop();
+        }
+    }
+
+    /**
+     * {@code GET /sessions/{id}/status} shows a worker's pending {@code fleet_ask} question, its
+     * {@code turnId} and its ticket only to the caller whose owner key created that delegation. An
+     * unnamed primary is held to the same rule: its owner key is {@code null}, which here does not
+     * match the named worker that created the delegation, so it sees none of the pending-ask
+     * fields either — the same as any other non-creating caller.
+     */
+    @Test
+    void restStatusGatesThePendingAskFieldsByTheDelegationsCreatorOwner() throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        AgentControl agents = new AgentControl(herdr);
+        Injector injector = new Injector(agents);
+        Rendezvous rendezvous = new Rendezvous();
+        MessageService messages = new MessageService(agents, injector, rendezvous);
+
+        Javalin creatorApp = startOnSharedService(messages, herdr, FakeHerdr.WORKER_PID); // -> term_a
+        Javalin otherWorkerApp = startOnSharedService(messages, herdr, 9001L);             // -> term_shell
+        Javalin primaryApp = startOnSharedService(messages, herdr, 999_999L);              // no pane -> primary
+        try {
+            ObjectMapper mapper = new ObjectMapper();
+            String ticket = messages.sendAsync("term_target", "task that asks", null,
+                    Principal.worker("term_a", FakeHerdr.WORKER_PID));
+            long deadline = System.currentTimeMillis() + 3000;
+            while (!rendezvous.isWaiting("term_target") && System.currentTimeMillis() < deadline) {
+                Thread.sleep(5);
+            }
+            assertTrue(rendezvous.isWaiting("term_target"), "sendAsync should have opened its rendezvous waiter");
+
+            CompletableFuture<MessageService.AskResult> ask = CompletableFuture.supplyAsync(
+                    () -> messages.ask("term_target", "which config file?", 5000));
+
+            MessageService.TaskView asking;
+            deadline = System.currentTimeMillis() + 3000;
+            do {
+                // the no-check overload: this is test plumbing waiting for ASKING, not the gate under test
+                asking = messages.poll(ticket);
+                Thread.sleep(5);
+            } while (asking.phase() != MessageService.Phase.ASKING && System.currentTimeMillis() < deadline);
+            assertEquals(MessageService.Phase.ASKING, asking.phase());
+            String turnId = asking.turnId();
+
+            JsonNode other = mapper.readTree(
+                    send(otherWorkerApp.port(), "GET", "/sessions/term_target/status", null, null).body());
+            assertEquals("idle", other.get("status").asText(), "the base status must still be shown");
+            assertFalse(other.has("question"), "a non-creating caller must not see the question: " + other);
+            assertFalse(other.has("turnId"), "a non-creating caller must not see the turnId: " + other);
+            assertFalse(other.has("ticket"), "a non-creating caller must not see the ticket: " + other);
+
+            JsonNode own = mapper.readTree(
+                    send(creatorApp.port(), "GET", "/sessions/term_target/status", null, null).body());
+            assertEquals("which config file?", own.get("question").asText(), "the creator must see the question");
+            assertEquals(turnId, own.get("turnId").asText(), "the creator must see the turnId");
+            assertEquals(ticket, own.get("ticket").asText(), "the creator must see the ticket");
+
+            JsonNode primary = mapper.readTree(
+                    send(primaryApp.port(), "GET", "/sessions/term_target/status", null, null).body());
+            assertEquals("idle", primary.get("status").asText(), "the base status must still be shown");
+            assertFalse(primary.has("question"),
+                    "an unnamed primary must not see a question on a delegation a named worker created: " + primary);
+            assertFalse(primary.has("turnId"), "a non-creating unnamed primary must not see the turnId: " + primary);
+            assertFalse(primary.has("ticket"), "a non-creating unnamed primary must not see the ticket: " + primary);
+
+            // Clean up the still-open ask so the background thread does not linger past the test.
+            CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                    () -> messages.answer(turnId, "config.yaml", 5000, "worker:term_a"));
+            assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+            deadline = System.currentTimeMillis() + 3000;
+            while (!rendezvous.isWaiting("term_target") && System.currentTimeMillis() < deadline) {
+                Thread.sleep(5);
+            }
+            assertTrue(rendezvous.resolve("term_target", "done"));
+            answer.get(5, TimeUnit.SECONDS);
+        } finally {
+            creatorApp.stop();
+            otherWorkerApp.stop();
+            primaryApp.stop();
+        }
+    }
+
+    /**
+     * {@code POST /sessions/{id}/message} with a {@code turnId} refuses a caller whose terminal
+     * did not open the turn, even when that caller otherwise holds ANSWER rights, and leaves the
+     * turn open for the real owner to resolve. Covers the REST adapter's ANSWER gate for an
+     * async-send ({@code wait:false}) delegation.
+     */
+    @Test
+    void restAnswerIsRefusedForADifferentCallerOnAnAsyncSendDelegationButTheRealOwnerSucceeds() throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        AgentControl agents = new AgentControl(herdr);
+        Injector injector = new Injector(agents);
+        Rendezvous rendezvous = new Rendezvous();
+        MessageService messages = new MessageService(agents, injector, rendezvous);
+
+        Javalin ownerApp = startOnSharedService(messages, herdr, FakeHerdr.WORKER_PID, Map.of("term_a", "lead-owner"));
+        Javalin attackerApp = startOnSharedService(messages, herdr, 9001L, Map.of("term_shell", "lead-attacker"));
+        try {
+            ObjectMapper mapper = new ObjectMapper();
+            HttpResponse<String> created = send(ownerApp.port(), "POST", "/sessions/term_target/message",
+                    "{\"content\":\"long task\",\"wait\":false}", null);
+            assertEquals(202, created.statusCode(), created.body());
+            String ticket = mapper.readTree(created.body()).path("ticket").asText(null);
+            assertNotNull(ticket, "the accepted response carried no ticket: " + created.body());
+
+            long deadline = System.currentTimeMillis() + 3000;
+            while (!rendezvous.isWaiting("term_target") && System.currentTimeMillis() < deadline) {
+                Thread.sleep(5);
+            }
+            assertTrue(rendezvous.isWaiting("term_target"), "the async send should have opened its rendezvous waiter");
+
+            CompletableFuture<MessageService.AskResult> ask = CompletableFuture.supplyAsync(
+                    () -> messages.ask("term_target", "which config file?", 5000));
+
+            MessageService.TaskView asking;
+            deadline = System.currentTimeMillis() + 3000;
+            do {
+                // the no-check overload: this is test plumbing waiting for ASKING, not the gate under test
+                asking = messages.poll(ticket);
+                Thread.sleep(5);
+            } while (asking.phase() != MessageService.Phase.ASKING && System.currentTimeMillis() < deadline);
+            assertEquals(MessageService.Phase.ASKING, asking.phase());
+            String turnId = asking.turnId();
+
+            HttpResponse<String> hijacked = send(attackerApp.port(), "POST", "/sessions/term_target/message",
+                    "{\"content\":\"hijack\",\"turnId\":\"" + turnId + "\"}", null);
+            assertEquals(403, hijacked.statusCode(), hijacked.body());
+            assertTrue(hijacked.body().contains("not_turn_owner"),
+                    "a different caller's answer must be refused as not_turn_owner: " + hijacked.body());
+            assertEquals("term_target", rendezvous.askSession(turnId),
+                    "a refused answer must leave the ask turn open");
+
+            CompletableFuture<HttpResponse<String>> owned = CompletableFuture.supplyAsync(() -> {
+                try {
+                    return send(ownerApp.port(), "POST", "/sessions/term_target/message",
+                            "{\"content\":\"config.yaml\",\"turnId\":\"" + turnId + "\"}", null);
+                } catch (Exception e) {
+                    throw new RuntimeException(e);
+                }
+            });
+            assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+            deadline = System.currentTimeMillis() + 3000;
+            while (!rendezvous.isWaiting("term_target") && System.currentTimeMillis() < deadline) {
+                Thread.sleep(5);
+            }
+            assertTrue(rendezvous.resolve("term_target", "done"));
+            HttpResponse<String> ownedResponse = owned.get(5, TimeUnit.SECONDS);
+            assertEquals(200, ownedResponse.statusCode(), ownedResponse.body());
+            assertEquals("done", mapper.readTree(ownedResponse.body()).path("reply").asText(null),
+                    "the real owner's answer must resolve the worker's turn");
+        } finally {
+            ownerApp.stop();
+            attackerApp.stop();
+        }
+    }
+
+    /**
+     * As above, but the delegation is a blocking send ({@code wait:true}) instead of a ticket —
+     * the owning caller's own HTTP request is the one that surfaces the worker's question and
+     * later carries the real answer. Covers the REST adapter's ANSWER gate for a blocking-send
+     * delegation.
+     */
+    @Test
+    void restAnswerIsRefusedForADifferentCallerOnABlockingSendDelegationButTheRealOwnerSucceeds() throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        AgentControl agents = new AgentControl(herdr);
+        Injector injector = new Injector(agents);
+        Rendezvous rendezvous = new Rendezvous();
+        MessageService messages = new MessageService(agents, injector, rendezvous);
+
+        Javalin ownerApp = startOnSharedService(messages, herdr, FakeHerdr.WORKER_PID, Map.of("term_a", "lead-owner"));
+        Javalin attackerApp = startOnSharedService(messages, herdr, 9001L, Map.of("term_shell", "lead-attacker"));
+        try {
+            ObjectMapper mapper = new ObjectMapper();
+            CompletableFuture<HttpResponse<String>> blocking = CompletableFuture.supplyAsync(() -> {
+                try {
+                    return send(ownerApp.port(), "POST", "/sessions/term_target/message",
+                            "{\"content\":\"long task\",\"wait\":true,\"timeoutMs\":5000}", null);
+                } catch (Exception e) {
+                    throw new RuntimeException(e);
+                }
+            });
+
+            long deadline = System.currentTimeMillis() + 3000;
+            while (!rendezvous.isWaiting("term_target") && System.currentTimeMillis() < deadline) {
+                Thread.sleep(5);
+            }
+            assertTrue(rendezvous.isWaiting("term_target"), "the blocking send should have opened its rendezvous waiter");
+
+            CompletableFuture<MessageService.AskResult> ask = CompletableFuture.supplyAsync(
+                    () -> messages.ask("term_target", "which config file?", 5000));
+
+            HttpResponse<String> questionResponse = blocking.get(5, TimeUnit.SECONDS);
+            assertEquals(202, questionResponse.statusCode(), questionResponse.body());
+            JsonNode question = mapper.readTree(questionResponse.body());
+            assertEquals("question", question.get("status").asText());
+            String turnId = question.get("turnId").asText();
+
+            HttpResponse<String> hijacked = send(attackerApp.port(), "POST", "/sessions/term_target/message",
+                    "{\"content\":\"hijack\",\"turnId\":\"" + turnId + "\"}", null);
+            assertEquals(403, hijacked.statusCode(), hijacked.body());
+            assertTrue(hijacked.body().contains("not_turn_owner"),
+                    "a different caller's answer must be refused as not_turn_owner: " + hijacked.body());
+            assertEquals("term_target", rendezvous.askSession(turnId),
+                    "a refused answer must leave the ask turn open");
+
+            CompletableFuture<HttpResponse<String>> owned = CompletableFuture.supplyAsync(() -> {
+                try {
+                    return send(ownerApp.port(), "POST", "/sessions/term_target/message",
+                            "{\"content\":\"config.yaml\",\"turnId\":\"" + turnId + "\"}", null);
+                } catch (Exception e) {
+                    throw new RuntimeException(e);
+                }
+            });
+            assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+            deadline = System.currentTimeMillis() + 3000;
+            while (!rendezvous.isWaiting("term_target") && System.currentTimeMillis() < deadline) {
+                Thread.sleep(5);
+            }
+            assertTrue(rendezvous.resolve("term_target", "done"));
+            HttpResponse<String> ownedResponse = owned.get(5, TimeUnit.SECONDS);
+            assertEquals(200, ownedResponse.statusCode(), ownedResponse.body());
+            assertEquals("done", mapper.readTree(ownedResponse.body()).path("reply").asText(null),
+                    "the real owner's answer must resolve the worker's turn");
+        } finally {
+            ownerApp.stop();
+            attackerApp.stop();
+        }
+    }
+
+    /**
+     * As {@link #start}, but shares {@code messages} and {@code herdr} across several app
+     * instances bound to different pids, each returned as its own started {@link Javalin} rather
+     * than through the shared {@code app} field, so several differently-resolved callers can
+     * poll the same ticket.
+     */
+    private Javalin startOnSharedService(MessageService messages, FakeHerdr herdr, long pid) {
+        return startOnSharedService(messages, herdr, pid, Map.of());
+    }
+
+    /**
+     * As {@link #startOnSharedService(MessageService, FakeHerdr, long)}, but {@code leadTerminals}
+     * resolves the given pid's terminal to a named lead (a caller with SEND permission) instead of
+     * a plain worker, for a test that needs a terminal-bearing caller able to create a ticket.
+     *
+     * <p>Every connecting pane not already claimed by {@code leadTerminals} is wired into the live
+     * roster as a spawned worker, so a caller's resolved role matches what its own test expects:
+     * a {@link Role#WORKER}, never the unconfigured-pane {@link Role#OBSERVER} floor a roster-less
+     * resolver would otherwise fall to.
+     */
+    private Javalin startOnSharedService(MessageService messages, FakeHerdr herdr, long pid,
+            Map<String, String> leadTerminals) {
+        FleetConfig.Profile wcfg = new FleetConfig.Profile(
+                "ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
+                "tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
+        AgentControl agents = new AgentControl(herdr);
+        ClaudeCodeLauncher workers = new ClaudeCodeLauncher(
+                agents, new WorkspaceControl(herdr), new SubscriptionGuard(Set.of("gx00.gw")),
+                Map.of(wcfg.profile(), wcfg), wcfg.profile(),
+                k -> "FLEETD_WORKER_TOKEN".equals(k) ? "tok-abc" : null);
+        SessionManager sessions = new SessionManager(workers, new FakeWorktrees());
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> pid);
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
+                () -> leadTerminals, new MemberRegistry(null),
+                t -> leadTerminals.containsKey(t) ? null : MemberRole.DEV, Map::of);
+        Metrics appMetrics = FleetMetrics.create(sessions, new dev.ltms.fleet.msg.InMemoryReplyInbox());
+
+        return new FleetApp(herdr, workers, sessions, messages, sessions.asPresence(), null,
+                callers, appMetrics).build().start("127.0.0.1", 0);
+    }
+
+    /**
+     * fleetd #669 Unit A: {@code POST /sessions/{id}/message} is two call shapes behind one route,
+     * mirroring {@code fleet_send}'s MCP-side split into {@link Authz.Action#SEND} and {@link
+     * Authz.Action#ANSWER} ({@code FleetMcp#sendAction}). The route never carries a {@code coordId}
+     * shape — that peer-lead route is MCP-only — so only these two apply here.
+     */
+    @Test
+    void theMessageRouteIsASendWithNoTurnIdAndAnAnswerWithOne() {
+        assertEquals(Authz.Action.SEND, FleetApp.routeAction("POST /sessions/{id}/message", null));
+        assertEquals(Authz.Action.SEND, FleetApp.routeAction("POST /sessions/{id}/message", "  "));
+        assertEquals(Authz.Action.ANSWER, FleetApp.routeAction("POST /sessions/{id}/message", "turn-1"));
+    }
+
+    /**
+     * fleetd #689: {@code answerGatePasses} is the second, conditional gate behind {@code
+     * sendMessage}'s coarse {@link Authz.Action#SEND} check. With the {@code ANSWER} grant denied,
+     * a {@code turnId}-bearing request is refused while a plain one still passes — and the denied
+     * permit is queried only for the {@code turnId} case, never for the plain one, which is what
+     * proves this is a genuinely separate, conditional check rather than the {@code SEND} check
+     * renamed or an unconditional call whose result is ignored. Flipping only the {@code ANSWER}
+     * grant to allowed then flips only the {@code turnId} shape's outcome.
+     */
+    @Test
+    void answerGatePassesOnlyWhenTurnIdAbsentOrAnswerGranted() {
+        List<Authz.Action> queried = new ArrayList<>();
+        Predicate<Authz.Action> denyAnswer = action -> {
+            queried.add(action);
+            return false;
+        };
+
+        assertFalse(FleetApp.answerGatePasses("turn-1", denyAnswer),
+                "ANSWER denied ⇒ the turnId shape is refused");
+        assertEquals(List.of(Authz.Action.ANSWER), queried,
+                "the ANSWER grant, specifically, must be the one consulted");
+
+        queried.clear();
+        assertTrue(FleetApp.answerGatePasses(null, denyAnswer),
+                "no turnId ⇒ the plain shape passes even though ANSWER is denied");
+        assertTrue(FleetApp.answerGatePasses("  ", denyAnswer), "a blank turnId is treated as absent");
+        assertEquals(List.of(), queried, "the plain shape must never consult the permit at all");
+
+        assertTrue(FleetApp.answerGatePasses("turn-1", action -> true),
+                "flipping only the ANSWER grant to allowed flips only the turnId shape's outcome");
+    }
+
     private static Set<String> routesTheServerRegisters() {
         try {
             String source = Files.readString(REST_SOURCE).lines()
@@ -147,6 +656,74 @@ class FleetAppAuthTest {
         }
     }
 
+    /**
+     * {@code permitsFor} is the exact decision {@link FleetApp#allow} makes, passing the real
+     * production classifier rather than a test-supplied one — built from an empty {@link
+     * CallerResolver}, so no terminal is recognised as a configured lead or collaborator and a
+     * collaborator's SEND is refused through the REST gate.
+     */
+    @Test
+    void aCollaboratorMayNotSendOverRestWithTheRealProductionClassifier() {
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(new FakeHerdr()), _ -> 700L);
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
+                Map::of, new MemberRegistry(null));
+        Principal collaborator = Principal.collaborator("ops", "term_collab", 700);
+        assertFalse(FleetApp.permitsFor(collaborator, Authz.Action.SEND, "term_lead",
+                        callers.knownLeadOrCollaborator()),
+                "no terminal is recognised as a lead or collaborator yet");
+    }
+
+    /**
+     * fleetd #669 Unit D: wires a real {@link CallerResolver} with a known lead and a known
+     * collaborator tab, and a spawned member's own terminal recognised by neither map. A
+     * collaborator's SEND reaches the known lead and the known collaborator, and is refused for the
+     * spawned member's terminal — over the REST route, not just the unit-level classifier, so a
+     * test covering only MCP cannot leave this route open.
+     */
+    @Test
+    void aCollaboratorMaySendToAKnownLeadOrCollaboratorButNotToASpawnedMembersTerminalOverRest() throws Exception {
+        int port = startWithRealClassifier(FakeHerdr.WORKER_PID,
+                Map.of("term_lead_known", "lead-x"), Map.of("term_a", "ops2"));
+
+        HttpResponse<String> toLead = send(port, "POST", "/sessions/term_lead_known/message",
+                "{\"content\":\"hi\",\"wait\":false}", null);
+        assertEquals(202, toLead.statusCode(), toLead.body());
+
+        HttpResponse<String> toSpawnedMembersTerminal = send(port, "POST", "/sessions/term_worker/message",
+                "{\"content\":\"hi\",\"wait\":false}", null);
+        assertEquals(403, toSpawnedMembersTerminal.statusCode(), toSpawnedMembersTerminal.body());
+    }
+
+    /**
+     * As {@link #start}, but with explicit lead/collaborator maps and no spawned-member roster, so
+     * a test can wire the real {@link CallerResolver#knownLeadOrCollaborator()} classifier instead
+     * of the default empty one.
+     */
+    private int startWithRealClassifier(long pid, Map<String, String> leadTerminals,
+                                        Map<String, String> collaboratorTerminals) {
+        FakeHerdr herdr = new FakeHerdr();
+        FleetConfig.Profile wcfg = new FleetConfig.Profile(
+                "ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
+                "tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
+        AgentControl agents = new AgentControl(herdr);
+        ClaudeCodeLauncher workers = new ClaudeCodeLauncher(
+                agents, new WorkspaceControl(herdr), new SubscriptionGuard(Set.of("gx00.gw")),
+                Map.of(wcfg.profile(), wcfg), wcfg.profile(),
+                k -> "FLEETD_WORKER_TOKEN".equals(k) ? "tok-abc" : null);
+        SessionManager sessions = new SessionManager(workers, new FakeWorktrees());
+        Injector injector = new Injector(agents);
+        MessageService messages = new MessageService(agents, injector, new Rendezvous());
+
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(herdr), _ -> pid);
+        CallerResolver callers = CallerResolver.withLeadsAndMembers(identity, false, null,
+                () -> leadTerminals, new MemberRegistry(null), t -> null, () -> collaboratorTerminals);
+        metrics = FleetMetrics.create(sessions, new dev.ltms.fleet.msg.InMemoryReplyInbox());
+
+        app = new FleetApp(herdr, workers, sessions, messages, sessions.asPresence(), null,
+                callers, metrics).build().start("127.0.0.1", 0);
+        return app.port();
+    }
+
     // --- loopback-trust: the caller is the primary -------------------------------------------
 
     @Test
@@ -192,6 +769,107 @@ class FleetAppAuthTest {
                 "draining an inbox is the primary's collection step");
     }
 
+    /**
+     * fleetd #669 Unit A: the {@code turnId} shape of {@code POST /sessions/{id}/message} maps to
+     * {@link Authz.Action#ANSWER}, not the plain {@link Authz.Action#SEND} the test above drives —
+     * a worker must stay refused on this shape too, exactly as it was refused on the one undivided
+     * action before the split.
+     */
+    @Test
+    void aWorkerMayNotAnswerAnotherSessionsBlockedQuestionOverRest() throws Exception {
+        int port = start(FakeHerdr.WORKER_PID, false, null);
+
+        assertEquals(403, send(port, "POST", "/sessions/term_b/message",
+                "{\"turnId\":\"turn-1\",\"content\":\"hi\"}", null).statusCode(),
+                "resolving another session's blocked question would be a worker escalating too");
+    }
+
+    /**
+     * fleetd #689: a caller refused the coarse {@link Authz.Action#SEND} grant is refused on
+     * {@code SEND} specifically, even on the {@code turnId}-bearing shape that otherwise raises
+     * the check to {@link Authz.Action#ANSWER} — proving {@code turnId} was never read from the
+     * body before the refusal (reading it would have changed which action is named in the 403).
+     * The same caller refused with no body at all gets the identical detail, which could not hold
+     * if the decision depended on anything read from the body. Control: a caller who IS granted
+     * reaches past the gate and the body is used normally.
+     */
+    @Test
+    void aDeniedCallerIsRefusedOnSendEvenWithATurnIdBodyAndNeverReadsTheBody() throws Exception {
+        int workerPort = start(FakeHerdr.WORKER_PID, false, null); // denied: not primary/architect
+
+        HttpResponse<String> withTurnId = send(workerPort, "POST", "/sessions/term_b/message",
+                "{\"turnId\":\"turn-1\",\"content\":\"hi\"}", null);
+        assertEquals(403, withTurnId.statusCode());
+        assertTrue(withTurnId.body().contains("may not SEND"),
+                "the SEND check must be the one that fired, not ANSWER — ANSWER would only be "
+                        + "reachable by having already read turnId out of the body");
+
+        HttpResponse<String> noBody = send(workerPort, "POST", "/sessions/term_b/message", null, null);
+        assertEquals(403, noBody.statusCode());
+        assertTrue(noBody.body().contains("may not SEND"),
+                "refused identically with no body at all — the refusal cannot depend on body content");
+
+        // Control: a primary IS granted SEND, so the same turnId body is read and acted on —
+        // reaching messages.answer, which reports this unknown turnId as a stale one.
+        int primaryPort = start(999_999, false, null);
+        HttpResponse<String> granted = send(primaryPort, "POST", "/sessions/term_b/message",
+                "{\"turnId\":\"turn-1\",\"content\":\"hi\"}", null);
+        assertEquals(409, granted.statusCode());
+        assertTrue(granted.body().contains("stale_turn"), "a granted caller's body IS read and acted on");
+    }
+
+    /**
+     * fleetd #689 (ticket comment 18353): the only place {@code sendMessage}'s call to {@code
+     * answerGatePasses} is observable is the audit trail — {@code allow()} logs an {@code
+     * "allowed"} entry for every granted action except {@code READ}/{@code METRICS}/{@code
+     * TASK_READ}, and {@code ANSWER} is none of those. A granted {@code turnId} request must
+     * therefore log both a {@code SEND} and an {@code ANSWER} entry; a granted plain request must
+     * log {@code SEND} alone. A unit test of the extracted helper pins the helper; this pins the
+     * call site — deleting the {@code answerGatePasses} call from {@code sendMessage} leaves the
+     * helper's own test green but turns this one red.
+     */
+    @Test
+    void aGrantedTurnIdRequestAuditsBothSendAndAnswerButAPlainRequestAuditsSendAlone() throws Exception {
+        int port = start(999_999, false, null); // primary: granted both SEND and ANSWER
+        ObjectMapper mapper = new ObjectMapper();
+
+        try (CapturedLog audit = CapturedLog.at("audit", Level.INFO)) {
+            send(port, "POST", "/sessions/term_b/message",
+                    "{\"turnId\":\"turn-1\",\"content\":\"hi\"}", null);
+
+            List<String> allowed = allowedActions(audit, mapper);
+            assertTrue(allowed.contains("SEND"),
+                    "a turnId request must still clear the coarse SEND grant first");
+            assertTrue(allowed.contains("ANSWER"),
+                    "a turnId request must ALSO clear the ANSWER grant — this is the call site itself");
+        }
+
+        try (CapturedLog audit = CapturedLog.at("audit", Level.INFO)) {
+            send(port, "POST", "/sessions/term_b/message",
+                    "{\"content\":\"hi\",\"timeoutMs\":50}", null);
+
+            List<String> allowed = allowedActions(audit, mapper);
+            assertEquals(List.of("SEND"), allowed,
+                    "a plain request must log SEND and nothing else — ANSWER is conditional on "
+                            + "turnId, not something every request happens to log");
+        }
+    }
+
+    private static List<String> allowedActions(CapturedLog audit, ObjectMapper mapper) {
+        return audit.events().stream()
+                .map(ILoggingEvent::getFormattedMessage)
+                .map(line -> {
+                    try {
+                        return mapper.readTree(line);
+                    } catch (Exception e) {
+                        throw new AssertionError("audit line is not valid JSON: " + line, e);
+                    }
+                })
+                .filter(n -> "allowed".equals(n.path("outcome").asText()))
+                .map(n -> n.path("action").asText())
+                .toList();
+    }
+
     // --- token mode ---------------------------------------------------------------------------
 
     @Test

fleetd/src/test/java/dev/ltms/fleet/rest/FleetAppTest.java

@@ -223,6 +223,37 @@ class FleetAppTest {
         assertTrue(body.has("detail"), res.body());
     }
 
+    @Test
+    void agentsReportsTheAgentsTabLabel() throws Exception {
+        FakeHerdr herdr = new FakeHerdr().withTab("w2", "w2:t7", "trinotes");
+        int port = start(herdr, "http://gx00.gw:8000", Set.of("gx00.gw"));
+
+        HttpResponse<String> res = req(port, "GET", "/agents");
+        assertEquals(200, res.statusCode(), res.body());
+        JsonNode agents = mapper.readTree(res.body()).get("agents");
+        assertEquals(1, agents.size());
+        assertEquals("sess-1111", agents.get(0).get("sessionId").asText());
+        assertEquals("trinotes", agents.get(0).get("label").asText());
+    }
+
+    /**
+     * The tab-label scan ({@code workspace.list}/{@code tab.list}) is decoration on top of
+     * {@code workers.list()}'s own agent roster, so its failure must not cost that roster: a row
+     * reports a {@code null} label instead, never the {@code herdr_error} envelope.
+     */
+    @Test
+    void agentsStillReportsTheRosterWhenTheLabelScanFails() throws Exception {
+        FakeHerdr herdr = new FakeHerdr().workspaceListFailsWith("unavailable");
+        int port = start(herdr, "http://gx00.gw:8000", Set.of("gx00.gw"));
+
+        HttpResponse<String> res = req(port, "GET", "/agents");
+        assertEquals(200, res.statusCode(), res.body());
+        JsonNode agents = mapper.readTree(res.body()).get("agents");
+        assertEquals(1, agents.size());
+        assertEquals("sess-1111", agents.get(0).get("sessionId").asText());
+        assertTrue(agents.get(0).get("label").isNull(), "a failed label scan must report a null label, not fail the roster: " + res.body());
+    }
+
     @Test
     void spawnWorkerLandsInOwnTabInWorkerSpaceAndInjectsBaseUrl() throws Exception {
         FakeHerdr herdr = new FakeHerdr();
@@ -675,6 +706,26 @@ class FleetAppTest {
         assertEquals(400, postMessage(port, "{}").statusCode());
     }
 
+    /**
+     * fleetd #689: a body that fails to parse is rejected with 400 before {@code turnId} is ever
+     * read from it, so it reaches neither {@code messages.answer} (which needs a {@code turnId})
+     * nor {@code messages.send} — confirmed here for {@code send} by the fake agent's idle status,
+     * which would otherwise make an immediate {@code agent.prompt} delivery observable.
+     */
+    @Test
+    void malformedBodyReturns400AndNeverReachesSendOrAnswer() throws Exception {
+        FakeHerdr herdr = new FakeHerdr().agentStatus("idle"); // idle ⇒ send would deliver right away if reached
+        int port = start(herdr, "http://gx00.gw:8000", Set.of("gx00.gw"));
+
+        HttpResponse<String> res = postMessage(port, "not json at all");
+        assertEquals(400, res.statusCode());
+        JsonNode err = mapper.readTree(res.body());
+        assertEquals("bad_request", err.get("error").asText());
+        assertEquals("body must be JSON", err.get("detail").asText());
+        assertFalse(herdr.called("agent.prompt"),
+                "a malformed body must never reach messages.send's delivery");
+    }
+
     @Test
     void sessionStatusReportsLiveAgentStatus() throws Exception {
         FakeHerdr herdr = new FakeHerdr().agentStatus("blocked");

fleetd/src/test/java/dev/ltms/fleet/session/PresenceRacingLauncher.java

@@ -0,0 +1,92 @@
+package dev.ltms.fleet.session;
+
+import dev.ltms.fleet.inject.MemberPresence;
+import dev.ltms.fleet.peer.Capability;
+import dev.ltms.fleet.peer.PeerHandle;
+import dev.ltms.fleet.peer.PeerLauncher;
+import dev.ltms.fleet.peer.SpawnRequest;
+import dev.ltms.fleet.placement.PlacementDecision;
+
+import java.util.List;
+import java.util.Set;
+
+/**
+ * {@link PeerLauncher} decorator that marks presence for a spawned terminal before returning its
+ * handle to the caller — the contact-then-register ordering fleetd #722 covers, where the
+ * terminal's MCP contact lands before {@link SessionManager#acquire} runs its own
+ * {@code registry.put}. The presence view is set after construction, via {@link #presence},
+ * because it is owned by the {@link SessionManager} this launcher is passed into.
+ */
+final class PresenceRacingLauncher implements PeerLauncher {
+
+    private final PeerLauncher delegate;
+    volatile MemberPresence presence;
+
+    PresenceRacingLauncher(PeerLauncher delegate) {
+        this.delegate = delegate;
+    }
+
+    @Override
+    public PeerHandle spawn(SpawnRequest req) {
+        PeerHandle handle = delegate.spawn(req);
+        presence.markPresent(handle.terminalId());
+        return handle;
+    }
+
+    @Override
+    public PeerHandle spawn(SpawnRequest req, PlacementDecision decision) {
+        PeerHandle handle = delegate.spawn(req, decision);
+        presence.markPresent(handle.terminalId());
+        return handle;
+    }
+
+    @Override
+    public Set<Capability> capabilities() {
+        return delegate.capabilities();
+    }
+
+    @Override
+    public Set<Capability> capabilitiesFor(String profileName) {
+        return delegate.capabilitiesFor(profileName);
+    }
+
+    @Override
+    public Set<String> profiles() {
+        return delegate.profiles();
+    }
+
+    @Override
+    public String defaultProfile() {
+        return delegate.defaultProfile();
+    }
+
+    @Override
+    public String effectiveCwd(SpawnRequest req) {
+        return delegate.effectiveCwd(req);
+    }
+
+    @Override
+    public List<String> parityOverlay(String profileName) {
+        return delegate.parityOverlay(profileName);
+    }
+
+    @Override
+    public List<?> list() {
+        return delegate.list();
+    }
+
+    @Override
+    public int reapOrphanWorkers() {
+        return delegate.reapOrphanWorkers();
+    }
+
+    @Override
+    public void stop(String id) {
+        delegate.stop(id);
+    }
+
+    @Override
+    public boolean clearContext(String id) {
+        return delegate.clearContext(id);
+    }
+}

fleetd/src/test/java/dev/ltms/fleet/session/SessionManagerTest.java

@@ -422,6 +422,53 @@ class SessionManagerTest {
                 "turn completion moves BUSY → DONE");
     }
 
+    // --- fleetd #722: registration and presence must reach READY whichever lands first --------
+
+    @Test
+    void registerThenContactReachesReadyForPlainSpawn() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr);
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
+
+        sessions.asPresence().markPresent(session.terminalId());
+
+        assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
+                "a presence contact that arrives after registration reaches READY");
+    }
+
+    @Test
+    void contactThenRegisterStillReachesReadyForPlainSpawn() {
+        // The racing launcher marks presence for the spawned terminal from inside spawn() —
+        // before SessionManager.acquire's own registry.put runs — modeling an MCP contact that
+        // lands in that window.
+        FakeHerdr herdr = new FakeHerdr();
+        FleetConfig.Profile cfg = new FleetConfig.Profile(
+                "ltms-local", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN",
+                List.of("ccs", "ltms-local"), "tab", "fleetd-workers",
+                "worker: {profile} #{n}", null, null, null);
+        ClaudeCodeLauncher workers = new ClaudeCodeLauncher(new AgentControl(herdr), new WorkspaceControl(herdr),
+                new SubscriptionGuard(Set.of("gx00.gw")), Map.of(cfg.profile(), cfg), cfg.profile(), _ -> null);
+        PresenceRacingLauncher race = new PresenceRacingLauncher(workers);
+        SessionManager sessions = new SessionManager(race);
+        race.presence = sessions.asPresence();
+
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
+
+        assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
+                "a presence contact that lands before registry.put must still reach READY");
+    }
+
+    @Test
+    void aTerminalNeverMarkedPresentStaysSpawningAfterRegistration() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr);
+
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
+
+        assertEquals(MemberSession.State.SPAWNING, sessions.get(session.paneId()).orElseThrow().state(),
+                "registration alone must not advance a terminal that was never marked present");
+    }
+
     @Test
     void releaseTearsDownWorkerAndRemovesFromRosterAndIsIdempotent() {
         FakeHerdr herdr = new FakeHerdr();
@@ -1405,6 +1452,105 @@ class SessionManagerTest {
                         + "dirty check threw");
     }
 
+    // --- fleetd #736: a release must forget the member's presence entry, not just its registry
+    // row ---------------------------------------------------------------------------------------
+
+    @Test
+    void releaseByPaneIdForgetsThePresenceEntry() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr);
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
+        String terminal = session.terminalId();
+        sessions.asPresence().markPresent(terminal);
+        assertTrue(sessions.asPresence().isPresent(terminal), "present before the release");
+
+        sessions.release(session.paneId());
+
+        assertFalse(sessions.asPresence().isPresent(terminal),
+                "release must forget the terminal's presence, not just remove its registry row");
+    }
+
+    @Test
+    void reapIdleForgetsThePresenceEntryToo() {
+        long[] clock = {0};
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr, () -> clock[0]);
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
+        String terminal = session.terminalId();
+        sessions.asPresence().markPresent(terminal);
+        assertTrue(sessions.asPresence().isPresent(terminal), "present before the reap");
+
+        clock[0] = 11;
+        assertEquals(1, sessions.reapIdle(10), "READY session past TTL is reaped");
+
+        assertFalse(sessions.asPresence().isPresent(terminal),
+                "the idle-reap release path (releaseIfCurrent) goes through the same teardown "
+                        + "funnel as an explicit release, so it must forget presence too");
+    }
+
+    @Test
+    void shutdownDrainAlsoForgetsThePresenceEntry() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr);
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller", "term_primary");
+        String terminal = session.terminalId();
+        sessions.asPresence().markPresent(terminal);
+        assertTrue(sessions.asPresence().isPresent(terminal), "present before the drain");
+
+        sessions.drainAll(TimeUnit.MILLISECONDS.toNanos(100));
+
+        assertFalse(sessions.asPresence().isPresent(terminal),
+                "a shutdown drain still ends the member's process, so presence must be cleared "
+                        + "exactly as it is for any other release cause");
+    }
+
+    @Test
+    void releaseOfAnUnknownPaneIdDoesNotThrow() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr);
+
+        assertDoesNotThrow(() -> sessions.release("no-such-pane"),
+                "releasing a pane id that was never registered must be a no-op, not a throw");
+    }
+
+    @Test
+    void releaseStillForgetsPresenceWhenDirtyCheckThrows() {
+        FakeHerdr herdr = new FakeHerdr();
+        RecordingWorktrees worktrees = new RecordingWorktrees();
+        SessionManager sessions = sessionManager(herdr, worktrees);
+        MemberSession s = sessions.acquire("ltms-local", null, "/caller/proj", null,
+                new WorktreeRequest("fleetd-736", null));
+        String terminal = s.terminalId();
+        sessions.asPresence().markPresent(terminal);
+        worktrees.failHasUncommittedWith(new WorktreeException("git status exited 128"));
+
+        assertDoesNotThrow(() -> sessions.release(s.paneId()),
+                "a throwing dirty check must not abort the release");
+
+        assertFalse(sessions.asPresence().isPresent(terminal),
+                "presence must be forgotten even when the dirty check throws, which pins the "
+                        + "forget call to the finally block that runs no matter what happened above");
+    }
+
+    @Test
+    void releaseLeavesADifferentStillLiveMembersPresenceUntouched() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr);
+        MemberSession released = sessions.acquire("ltms-local", null, "/caller/a", "ownerA");
+        MemberSession stillLive = sessions.acquire("ltms-local", null, "/caller/b", "ownerB");
+        sessions.asPresence().markPresent(released.terminalId());
+        sessions.asPresence().markPresent(stillLive.terminalId());
+        assertTrue(sessions.asPresence().isPresent(stillLive.terminalId()),
+                "present before the release of the other member");
+
+        sessions.release(released.paneId());
+
+        assertFalse(sessions.asPresence().isPresent(released.terminalId()),
+                "the released terminal is forgotten");
+        assertTrue(sessions.asPresence().isPresent(stillLive.terminalId()),
+                "a still-live member's presence must survive an unrelated release");
+    }
+
     // --- fleetd #316: the dirty check must be re-taken after the worker is stopped, not trusted
     // stale from before it ------------------------------------------------------------------------
 
@@ -2436,4 +2582,160 @@ class SessionManagerTest {
             return "threw:" + e.getClass().getName() + ":" + e.getMessage();
         }
     }
+
+    // --- fleetd #702: spawnedMemberRole must still answer for a pane mid-teardown ----------------
+
+    /**
+     * A {@link Worktrees} test double whose {@code hasUncommitted} runs an injected hook before
+     * answering. This is what lets a test resolve a releasing pane's terminal from inside the
+     * window {@link SessionManager#release} opens between removing the registry entry and
+     * actually stopping the pane — the hook runs synchronously on the release call's own thread,
+     * at the exact point {@code release} shells out to {@code git status}, so there is no sleep
+     * and no race to land in it.
+     */
+    private static final class HookedWorktrees implements Worktrees {
+        private Runnable hook;
+        private boolean dirty = false;
+
+        HookedWorktrees onHasUncommitted(Runnable hook) {
+            this.hook = hook;
+            return this;
+        }
+
+        @Override
+        public String add(String repoRoot, String branch, String baseRef) {
+            return "/wt/" + branch.replace('/', '_');
+        }
+
+        @Override
+        public void remove(String repoRoot, String worktreePath) {
+        }
+
+        @Override
+        public void deleteBranch(String repoRoot, String branch) {
+        }
+
+        @Override
+        public boolean hasUncommitted(String worktreePath) {
+            if (hook != null) {
+                hook.run();
+            }
+            return dirty;
+        }
+
+        @Override
+        public void overlayParity(String repoRoot, String worktreePath, List<String> overlay) {
+        }
+
+        @Override
+        public String repoRoot(String cwd) {
+            return "/repo";
+        }
+
+        @Override
+        public java.util.Optional<String> snapshot(String worktreePath, String branch, String message) {
+            return java.util.Optional.empty();
+        }
+
+        @Override
+        public WipRefStats wipRefs(String repoRoot) {
+            return new WipRefStats(0, 0L);
+        }
+
+        @Override
+        public int pruneWipRefs(String repoRoot, long minAgeMillis) {
+            return 0;
+        }
+
+        @Override
+        public void shareWithGroup(String repoRoot, String worktreePath) {
+        }
+    }
+
+    @Test
+    void spawnedMemberRoleResolvesTheLiveRegisteredRole() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr);
+        MemberSession s = sessions.acquire("ltms-local", null, "/caller", null);
+
+        assertEquals(s.role(), sessions.spawnedMemberRole(s.terminalId()),
+                "a registered session resolves to its own role");
+        assertNull(sessions.spawnedMemberRole("term_unknown"),
+                "a terminal with no session at all resolves to null");
+    }
+
+    @Test
+    void spawnedMemberRoleIsNullOnceReleaseFullyCompletes() {
+        FakeHerdr herdr = new FakeHerdr();
+        SessionManager sessions = sessionManager(herdr, new HookedWorktrees());
+        MemberSession s = sessions.acquire("ltms-local", null, "/caller/proj", null,
+                new WorktreeRequest("fleetd-702a", null));
+
+        sessions.release(s.paneId());
+
+        assertNull(sessions.spawnedMemberRole(s.terminalId()),
+                "once release has fully finished, the terminal is neither registered nor releasing");
+    }
+
+    @Test
+    void spawnedMemberRoleStillAnswersBetweenTheRegistryRemovalAndThePaneStop() {
+        FakeHerdr herdr = new FakeHerdr();
+        java.util.concurrent.atomic.AtomicReference<MemberRole> duringWindow = new java.util.concurrent.atomic.AtomicReference<>();
+        HookedWorktrees worktrees = new HookedWorktrees();
+        SessionManager sessions = sessionManager(herdr, worktrees);
+        MemberSession s = sessions.acquire("ltms-local", null, "/caller/proj", null,
+                new WorktreeRequest("fleetd-702b", null));
+        worktrees.onHasUncommitted(() -> {
+            // This runs from INSIDE release()'s git-status shell-out: the registry entry is
+            // already gone, but the pane has not stopped yet — a real call landing inside the
+            // exact window fleetd #702 reports, so no sleep and no race is needed to reach it.
+            assertTrue(sessions.get(s.paneId()).isEmpty(),
+                    "sanity: the registry entry is already gone at this point");
+            duringWindow.set(sessions.spawnedMemberRole(s.terminalId()));
+        });
+
+        sessions.release(s.paneId());
+
+        assertEquals(s.role(), duringWindow.get(),
+                "spawnedMemberRole must still answer the live role while the pane is mid-teardown, "
+                        + "not only while the session is still in the registry");
+        assertNull(sessions.spawnedMemberRole(s.terminalId()),
+                "and once release has fully finished, the window is closed too");
+    }
+
+    /**
+     * {@code Releasing.enter} and {@code Releasing.leave} are called directly here, never through
+     * {@link SessionManager#release} or {@link SessionManager#spawnedMemberRole}, so a test naming
+     * one of them exercises only that one — a regression in the other can never hide behind it.
+     */
+    @Test
+    void releasingLeaveStepsDownADepthGreaterThanOneInsteadOfRemovingIt() {
+        SessionManager.Releasing depthTwo = new SessionManager.Releasing(2, "term_a", MemberRole.DEV);
+
+        SessionManager.Releasing afterLeave = depthTwo.leave();
+
+        assertNotNull(afterLeave,
+                "depth 2 means another release of the SAME pane is still mid-teardown; leave() must "
+                        + "step the depth down, never remove the marker outright — removing it here "
+                        + "is what a plain Set would do, and would reopen the window the still-in-"
+                        + "flight release is relying on staying closed");
+        assertEquals(1, afterLeave.depth());
+        assertEquals("term_a", afterLeave.terminalId());
+        assertEquals(MemberRole.DEV, afterLeave.role());
+    }
+
+    @Test
+    void releasingEnterPreservesThePriorTerminalWhenTheOverlappingCallHasNoSessionOfItsOwn() {
+        SessionManager.Releasing prior = new SessionManager.Releasing(1, "term_a", MemberRole.DEV);
+
+        SessionManager.Releasing afterEnter = SessionManager.Releasing.enter(prior, null);
+
+        assertEquals(2, afterEnter.depth(), "depth still increments whether or not this enter knows its session");
+        assertEquals("term_a", afterEnter.terminalId(),
+                "an overlapping release that finds the registry entry already gone has no session of "
+                        + "its own to pass as known, and must not blank out the terminal the first "
+                        + "call already recorded — that terminal is what spawnedMemberRole matches "
+                        + "against");
+        assertEquals(MemberRole.DEV, afterEnter.role());
+    }
 }

fleetd/src/test/java/dev/ltms/fleet/session/WorktreeSessionManagerTest.java

@@ -159,6 +159,40 @@ class WorktreeSessionManagerTest {
         assertEquals(expectedPath, s.cwd(), "session cwd is the worktree path");
     }
 
+    // --- fleetd #722: registration and presence must reach READY whichever lands first --------
+
+    @Test
+    void registerThenContactReachesReadyForWorktreeSpawn() {
+        FakeHerdr herdr = new FakeHerdr();
+        FakeWorktrees worktrees = new FakeWorktrees().withRepoRoot("/repo").withPrefix("/wt");
+        SessionManager sessions = new SessionManager(workerService(herdr), worktrees);
+
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller/proj", "term_primary",
+                new WorktreeRequest("cb-722", null));
+        sessions.asPresence().markPresent(session.terminalId());
+
+        assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
+                "a presence contact that arrives after worktree registration reaches READY");
+    }
+
+    @Test
+    void contactThenRegisterStillReachesReadyForWorktreeSpawn() {
+        // The racing launcher marks presence for the spawned terminal from inside spawn() —
+        // before SessionManager.acquireWithWorktree's own registry.put runs — modeling an MCP
+        // contact that lands in that window.
+        FakeHerdr herdr = new FakeHerdr();
+        FakeWorktrees worktrees = new FakeWorktrees().withRepoRoot("/repo").withPrefix("/wt");
+        PresenceRacingLauncher race = new PresenceRacingLauncher(workerService(herdr));
+        SessionManager sessions = new SessionManager(race, worktrees);
+        race.presence = sessions.asPresence();
+
+        MemberSession session = sessions.acquire("ltms-local", null, "/caller/proj", "term_primary",
+                new WorktreeRequest("cb-722", null));
+
+        assertEquals(MemberSession.State.READY, sessions.get(session.paneId()).orElseThrow().state(),
+                "a presence contact that lands before worktree registration must still reach READY");
+    }
+
     @Test
     void worktreeArchitectAcquireAlsoBindsItsSlot() {
         FakeHerdr herdr = new FakeHerdr();

scripts/config-edit.sh

@@ -213,6 +213,17 @@ map_masked_lines() {
   done < "$file"
 }
 
+# Masks every `scheme://user:pass@host` userinfo on one line of text, replacing just that
+# userinfo with `<redacted>` and leaving the rest of the line untouched, byte for byte. The
+# pattern stops at the first `/`, whitespace, or `@` reached after `://` — a URI's userinfo
+# component cannot contain any of those three characters — so a URI with no userinfo, followed
+# later on the same line by an unrelated `@`, never matches. The `g` flag matters: a line can
+# carry more than one URI. Shared by every caller that prints a line which may hold a
+# credentialed URI, so the bound lives in exactly one place.
+mask_url_userinfo() {
+  printf '%s\n' "$1" | sed -E 's#://[^@/[:space:]]*@#://<redacted>@#g'
+}
+
 redact() {
   local old_file="$1" new_file="$2"
   local line prefix content indent lead key old_line=0 new_line=0 in_hunk=0
@@ -270,7 +281,7 @@ redact() {
         continue
       fi
     fi
-    printf '%s\n' "$line" | sed -E 's#://[^@]*@#://<redacted>@#g'
+    mask_url_userinfo "$line"
   done
   [ "$saved_nocasematch" = 1 ] || shopt -u nocasematch
 }
@@ -580,6 +591,12 @@ install_candidate() {
 
 # -------------------------------------------------------------------------------- the report path
 #
+# Masks basic-auth userinfo (scheme://user:pass@host) in a daemon verdict line before it reaches
+# the terminal.
+mask_verdict_userinfo() {
+  mask_url_userinfo "$1"
+}
+
 # Prints the literal command the operator (or a test) can run to restore the backup by hand — the
 # absolute path to THIS script plus the overrides actually in force, so it works from any cwd.
 restore_command_line() {
@@ -599,8 +616,8 @@ restore_and_confirm() {
   ok "restored from $backup"
   if wait_for_verdict "$LOG" "$mark2" "$WAIT_SECONDS"; then
     case "$VERDICT_KIND" in
-      refused) warn "the RESTORE was also refused by the daemon: $VERDICT_LINE" ;;
-      *)       ok "restore confirmed: $VERDICT_LINE" ;;
+      refused) warn "the RESTORE was also refused by the daemon: $(mask_verdict_userinfo "$VERDICT_LINE")" ;;
+      *)       ok "restore confirmed: $(mask_verdict_userinfo "$VERDICT_LINE")" ;;
     esac
   else
     warn "the restore is on disk, but no confirming verdict line appeared within ${WAIT_SECONDS}s"
@@ -616,7 +633,7 @@ report_outcome() {
 
   say "waiting for the daemon's verdict (up to ${WAIT_SECONDS}s)"
   if wait_for_verdict "$LOG" "$mark" "$WAIT_SECONDS"; then
-    kind="$VERDICT_KIND"; line="$VERDICT_LINE"
+    kind="$VERDICT_KIND"; line="$(mask_verdict_userinfo "$VERDICT_LINE")"
   else
     kind="none"
   fi
@@ -673,7 +690,7 @@ check_mode() {
   local verdict
   verdict="$(last_verdict_line "$LOG")"
   if [ -n "$verdict" ]; then
-    ok "last verdict in log: $verdict"
+    ok "last verdict in log: $(mask_verdict_userinfo "$verdict")"
   else
     warn "no reload verdict line found in $LOG"
   fi

scripts/redeploy-fleetd.sh

@@ -1507,6 +1507,7 @@ else
   grep -E ' (ERROR|SEVERE) ' "$FRESH_LOG" | tail -5 | sed 's/^/   /'
 fi
 echo
-echo "   Next: call fleet_whoami and confirm it still answers 'primary'. A lead whose tab label"
-echo "   no longer matches fleet.leaders.*.tab is demoted to worker and refuses orchestration."
+echo "   Next: call fleet_whoami and confirm it still answers 'primary'. A lead is found by its"
+echo "   tab being labelled 'lead' AND sitting in fleet.leaders.<name>.workspace; if either stops"
+echo "   matching, the lead is demoted to worker and refuses orchestration."
 echo

scripts/test-config-edit.sh

@@ -233,6 +233,66 @@ test_redaction_holds() {
   assert_contains "weight" "$RUN_OUTPUT" "a diff must have been demonstrably printed at all"
 }
 
+# redact()'s key-name filter only inspects the KEY, so a diff line whose key does not match
+# TOKEN|SECRET|PASSWORD|PASSWD|PASSPHRASE|CREDENTIAL|URI|_KEY still reaches the final userinfo
+# sed even when its VALUE holds a credentialed URI. "note" is not a sensitive key name, so this
+# line must fall all the way through to that sed, not the earlier whole-value branch. The
+# trailing prose on both sides of the userinfo is a positive control: it proves the line reached
+# the userinfo sed (which touches only the userinfo) rather than the earlier branch (which would
+# have replaced the whole value with a bare "<redacted>" and dropped the prose).
+test_diff_line_userinfo_is_masked_with_positive_control() {
+  local dir
+  dir="$(new_fixture)"
+
+  start_run "$dir" 5 --set '.profiles.sonnet.note=see amqp://alice:wonderland@rabbit.local:5672/vhost for details'
+  sleep 1
+  printf 'config reloaded\n' >> "$dir/fleetd.out"
+  collect_run "$dir"
+
+  assert_equals 0 "$RUN_RC" "diff-userinfo-case reload exit code"
+  assert_not_contains "alice:wonderland" "$RUN_OUTPUT" "the userinfo must never reach the output"
+  assert_contains "amqp://<redacted>@rabbit.local:5672/vhost" "$RUN_OUTPUT" \
+    "the userinfo must be MASKED, not deleted — the rest of the value must survive"
+  assert_contains "note:" "$RUN_OUTPUT" "the key name must still reach the output"
+  assert_contains "see " "$RUN_OUTPUT" "prose BEFORE the userinfo must still reach the output"
+  assert_contains "for details" "$RUN_OUTPUT" "prose AFTER the userinfo must still reach the output"
+}
+
+# A diff line can hold a URL with no userinfo, followed later on the same line by an unrelated @
+# (free text in a string value, for example an email address). The line must pass through the
+# userinfo sed byte for byte: the match must stop at the end of the URL and must not treat the
+# later @ as a second userinfo delimiter.
+test_diff_line_uri_without_userinfo_survives_a_later_at_sign() {
+  local dir
+  dir="$(new_fixture)"
+
+  start_run "$dir" 5 --set '.profiles.sonnet.note2=see https://docs.local/guide and mail ops@example.com'
+  sleep 1
+  printf 'config reloaded\n' >> "$dir/fleetd.out"
+  collect_run "$dir"
+
+  assert_equals 0 "$RUN_RC" "diff-no-userinfo-with-later-at-sign reload exit code"
+  assert_contains "note2: see https://docs.local/guide and mail ops@example.com" "$RUN_OUTPUT" \
+    "a URL with no userinfo plus a later @ on the same line must pass through byte for byte"
+}
+
+# Two credentialed URIs on one diff line must both be masked — the g flag matters.
+test_diff_line_masks_multiple_userinfo_with_g_flag() {
+  local dir
+  dir="$(new_fixture)"
+
+  start_run "$dir" 5 --set '.profiles.sonnet.note3=amqp://u1:p1@host1/vhost1 and amqp://u2:p2@host2/vhost2'
+  sleep 1
+  printf 'config reloaded\n' >> "$dir/fleetd.out"
+  collect_run "$dir"
+
+  assert_equals 0 "$RUN_RC" "diff-two-userinfo-on-one-line reload exit code"
+  assert_not_contains "u1:p1" "$RUN_OUTPUT" "the first userinfo must never reach the output"
+  assert_not_contains "u2:p2" "$RUN_OUTPUT" "the second userinfo must never reach the output"
+  assert_contains "amqp://<redacted>@host1/vhost1" "$RUN_OUTPUT" "the first URI must be masked"
+  assert_contains "amqp://<redacted>@host2/vhost2" "$RUN_OUTPUT" "the second URI must be masked"
+}
+
 # ------------------------------------------------------- acceptance criterion 9: forgotten value
 # `--set .a.b=` is a plausible typo (the value simply forgotten), and it must be refused outright
 # rather than silently nulling the field — a null numeric field falls back to its default, which
@@ -629,6 +689,65 @@ test_refusal_shape_from_parse_failure_wording_is_recognised() {
   assert_equals 4 "$RUN_RC" "the parse-failure refusal shape must also exit 4, not be read as silence"
 }
 
+# A verdict line carrying a credentialed URI has its userinfo masked, with a positive control
+# proving the rest of the line still reaches the output unchanged.
+test_verdict_userinfo_is_masked_with_positive_control() {
+  local dir
+  dir="$(new_fixture)"
+
+  start_run "$dir" 5 --set '.profiles.sonnet.weight=4'
+  sleep 1
+  printf 'config reload from %s refused, keeping the running config: refusing to start: malformed pattern — profiles.local.errorPattern ("amqp://user:hunter2@host/vhost"): Unclosed character class near index 8\n' \
+    "$dir/fleetd.yaml" >> "$dir/fleetd.out"
+  collect_run "$dir"
+
+  assert_equals 4 "$RUN_RC" "refusal-with-userinfo exit code"
+  assert_not_contains "user:hunter2" "$RUN_OUTPUT" "the userinfo must never reach the output"
+  assert_contains "amqp://<redacted>@host/vhost" "$RUN_OUTPUT" \
+    "the userinfo must be MASKED, not deleted — the rest of the quoted value must survive"
+  # Positive control: the diagnostic prose on both sides of the userinfo must still reach the
+  # output. Without this, a mutant that drops the whole verdict line would pass identically.
+  assert_contains "malformed pattern" "$RUN_OUTPUT" "prose BEFORE the userinfo must still reach the output"
+  assert_contains "Unclosed character class near index 8" "$RUN_OUTPUT" \
+    "prose AFTER the userinfo must still reach the output"
+}
+
+# An ordinary refusal line quotes the offending pattern, not a credential, and must survive byte
+# for byte: the rewrite is scoped to userinfo only, and the quoted pattern is the detail an
+# operator needs to fix the refusal.
+test_ordinary_refusal_line_passes_through_unchanged() {
+  local dir real_line
+  dir="$(new_fixture)"
+  real_line="config reload from $dir/fleetd.yaml refused, keeping the running config: refusing to start: malformed pattern — profiles.local.errorPattern (\"[unclosed\"): Unclosed character class near index 8"
+
+  start_run "$dir" 5 --set '.profiles.sonnet.weight=4'
+  sleep 1
+  printf '%s\n' "$real_line" >> "$dir/fleetd.out"
+  collect_run "$dir"
+
+  assert_equals 4 "$RUN_RC" "ordinary refusal exit code"
+  assert_contains "$real_line" "$RUN_OUTPUT" \
+    "an ordinary refusal with no userinfo must pass through byte for byte, unchanged"
+}
+
+# A verdict line can hold a URI with NO userinfo and a later, unrelated @ further on in the same
+# line (an email address in diagnostic prose, for example). The rewrite must stop at the end of
+# the URI and must not treat the later @ as a second userinfo delimiter.
+test_uri_without_userinfo_survives_a_later_at_sign() {
+  local dir real_line
+  dir="$(new_fixture)"
+  real_line="config reload from $dir/fleetd.yaml refused, keeping the running config: broker.uri amqp://broker.local/vhost unreachable, contact ops@example.com"
+
+  start_run "$dir" 5 --set '.profiles.sonnet.weight=4'
+  sleep 1
+  printf '%s\n' "$real_line" >> "$dir/fleetd.out"
+  collect_run "$dir"
+
+  assert_equals 4 "$RUN_RC" "no-userinfo-with-later-at-sign exit code"
+  assert_contains "$real_line" "$RUN_OUTPUT" \
+    "a URI with no userinfo plus a later @ in the same line must pass through byte for byte"
+}
+
 # --set runs yq over the whole candidate. It warns when that changes more lines than the requested
 # pairs, but a simple file with only the intended changed line must stay quiet.
 new_fixture_reformat_sensitive() {
@@ -694,6 +813,12 @@ echo "== acceptance criterion 6: the marker works =="
 test_marker_skips_lines_before_it
 echo "== acceptance criterion 7 (+13: redaction is proven to have run) =="
 test_redaction_holds
+echo "== fleetd #692: a diff line's userinfo is masked, rest of the value survives =="
+test_diff_line_userinfo_is_masked_with_positive_control
+echo "== fleetd #692: a diff line's URI with no userinfo survives a later @ in the line =="
+test_diff_line_uri_without_userinfo_survives_a_later_at_sign
+echo "== fleetd #692: two userinfo URIs on one diff line are both masked =="
+test_diff_line_masks_multiple_userinfo_with_g_flag
 echo "== acceptance criterion 9: a forgotten value refuses and installs nothing =="
 test_forgotten_value_refuses_and_installs_nothing
 echo "== acceptance criterion 10: an explicit clear writes a bare null =="
@@ -720,6 +845,12 @@ echo "== extra: --check is read-only and always exits 0 =="
 test_check_is_read_only_and_exits_zero
 echo "== extra: the parse-failure refusal shape is also recognised =="
 test_refusal_shape_from_parse_failure_wording_is_recognised
+echo "== verdict-redaction criteria 2+3: verdict userinfo is masked, rest of line survives =="
+test_verdict_userinfo_is_masked_with_positive_control
+echo "== verdict-redaction criterion 4: an ordinary refusal passes through unchanged =="
+test_ordinary_refusal_line_passes_through_unchanged
+echo "== fleetd #638: a URI with no userinfo survives a later @ in the same line =="
+test_uri_without_userinfo_survives_a_later_at_sign
 echo "== acceptance criterion 17: --set warns about yq formatting churn =="
 test_set_warns_when_yq_reformats_extra_lines
 echo "== acceptance criterion 18: --set stays quiet without formatting churn =="