CB-619 / fleetd #123: refuse an architect spawn with no matching slot

An explicit-profile spawn bypasses role-pool placement (CompositePeerLauncher
only constrains an UNQUALIFIED spawn to fleet.<role>), so it was the one path
that could ask for role=architect on a profile no architect slot carries.
MemberRegistry silently held the session as a plain worker while GET /members
still reported the requested "architect" and only fleet_whoami (which reads
live bindings, not the request) told the truth.

- MemberLifecycle.requireSlotFor(role, profile): refuses the acquire before
  anything spawns when no configured architect slot carries the profile,
  naming the role, the profile, and the pools that do carry it. No-op for
  dev/reviewer, which are placement candidates only, never a live identity
  binding — refusing a profile mismatch there would break the documented
  fleet_spawn{profile:"opus"} (role defaults to dev) flow.
- MemberLifecycle.acquired(...) now returns the role the session actually
  holds, so a residual race (a slot exists but every instance is already
  bound to a different terminal) still falls back to dev honestly instead of
  lying — this case logs at WARN (was INFO), naming profile and terminal.
- SessionManager now records the role acquired() returns on MemberSession,
  never the requested role, so GET /members and fleet_list can no longer
  report a role the member does not hold; no changes needed to memberView/
  rosterView, which just read session.role().

An architect's identity IS the slot it is bound to — binding a role with no
slot to bind means inventing an identity out of nothing, which is the quiet
failure the whole role system exists to prevent.

Tests: SessionManagerTest and FleetMcpTest each drive a real spawn through
FleetMcp.spawn -> SessionManager.acquire -> the real ClaudeCodeLauncher (via
FakeHerdr), then assert on GET /members and fleet_whoami for that same
session — not on MemberRegistry.bind directly (fleetd issue #113's mistake).

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

@@ -7,15 +7,43 @@ public interface MemberLifecycle {
 
     MemberLifecycle NONE = new MemberLifecycle() {
         @Override
-        public void acquired(MemberRole role, String profile, String terminal) {
+        public MemberRole acquired(MemberRole role, String profile, String terminal) {
+            return role; // no registry configured — nothing to bind against, so the request stands
         }
 
         @Override
         public void released(String terminal) {
         }
+
+        @Override
+        public void requireSlotFor(MemberRole role, String profile) {
+            // no registry configured — nothing to validate against, so nothing is refused
+        }
     };
 
-    void acquired(MemberRole role, String profile, String terminal);
+    /**
+     * Try to bind a newly spawned {@code terminal} into the role it was granted.
+     *
+     * @return the role this session actually holds: {@code role} unchanged for a role with no
+     *         live slot-binding semantics (dev, reviewer), or when the bind succeeded; a fallback
+     *         role — never {@code role} — when a slot-bound role (architect) could not be bound.
+     *         Callers must record THIS value on the session, never the requested {@code role}, so
+     *         a later roster read never reports a role the session does not hold (CB-619). In
+     *         normal operation this fallback should not happen once {@link #requireSlotFor} has
+     *         refused every unbindable spawn upfront — but a slot can still be lost between that
+     *         check and this call to a concurrent spawn racing for the same slot, so the honest
+     *         answer is still needed here too.
+     */
+    MemberRole acquired(MemberRole role, String profile, String terminal);
 
     void released(String terminal);
+
+    /**
+     * Refuse an acquire before anything spawns when {@code role} requires a live slot binding and
+     * no configured slot carries {@code profile} (CB-619 / fleetd #123). A no-op for a role with
+     * no slot-binding semantics.
+     *
+     * @throws IllegalArgumentException naming the role, the profile, and the pools that do carry it
+     */
+    void requireSlotFor(MemberRole role, String profile);
 }

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

@@ -8,6 +8,7 @@ import org.slf4j.LoggerFactory;
 import java.util.Collections;
 import java.util.HashMap;
 import java.util.LinkedHashMap;
+import java.util.List;
 import java.util.Map;
 import java.util.Objects;
 
@@ -206,19 +207,72 @@ public final class MemberRegistry implements MemberLifecycle {
      *
      * <p>The role check is lifecycle policy. {@link CallerResolver} repeats it when resolving a
      * binding, so a later lifecycle regression cannot turn a worker into an architect.
+     *
+     * <p>CB-619 / fleetd #123: the return value is the role this session actually holds, and the
+     * caller is required to record THAT — never the requested {@code role} — on the session. Before
+     * this fix the caller kept the requested role regardless of whether the bind below succeeded, so
+     * a demoted session's {@code GET /members} row still said {@code "architect"} while
+     * {@code fleet_whoami} (which reads the live binding, not the request) correctly said
+     * {@code "worker"} — three sources of truth that disagreed about one live member, silently.
      */
     @Override
-    public void acquired(MemberRole role, String profile, String terminal) {
+    public MemberRole acquired(MemberRole role, String profile, String terminal) {
         if (role != MemberRole.ARCHITECT || terminal == null || terminal.isBlank()) {
-            return;
+            return role;
         }
         // slotsFor preserves definition order, so duplicate-profile slots use the first free one.
         for (Entry entry : slotsFor(MemberRole.ARCHITECT).values()) {
             if (Objects.equals(profile, entry.profile()) && bind(entry.key(), terminal)) {
-                return;
+                return MemberRole.ARCHITECT;
             }
         }
-        log.info("member slot: no free architect slot for profile={}; session remains a worker", profile);
+        // fleetd #123: at least WARN — a role downgrade that the roster must now also reflect is
+        // not routine bookkeeping. requireSlotFor already refuses the config-gap case (no slot at
+        // all carries this profile) before a process ever spawns; reaching here means the config DID
+        // carry a matching slot but every one of them was already bound to a different terminal — a
+        // race this pre-spawn check cannot close on its own (see requireSlotFor's javadoc).
+        log.warn("member slot: no free architect slot for profile={} terminal={}; holding the session "
+                + "as {} instead of the architect it asked for — every configured slot for this "
+                + "profile is already bound to a different terminal", profile, terminal,
+                MemberRole.DEV.wireName());
+        return MemberRole.DEV;
+    }
+
+    /**
+     * CB-619 / fleetd #123: refuse an architect acquire before anything spawns when no configured
+     * slot carries {@code profile} — the config-gap case from the original defect report (a spawn
+     * asked for {@code role=architect, profile=sonnet}, and {@code fleet.architects} carried only
+     * {@code opus} and {@code sol}). A dev/reviewer acquire is always a no-op: those pools are
+     * placement candidates only (see {@code CompositePeerLauncher}), never a live identity binding,
+     * so there is nothing here to refuse — an explicit profile outside the pool for those roles is a
+     * documented operator override, not a defect.
+     *
+     * <p>This closes the config-gap case, not the live-capacity case: a profile that DOES carry a
+     * slot can still lose the race to a concurrent spawn between this check and the actual
+     * {@link #bind}, which is why {@link #acquired} must still answer honestly even after this
+     * check has passed.
+     */
+    @Override
+    public void requireSlotFor(MemberRole role, String profile) {
+        if (role != MemberRole.ARCHITECT) {
+            return;
+        }
+        boolean hasSlot = slotsFor(MemberRole.ARCHITECT).values().stream()
+                .anyMatch(e -> Objects.equals(profile, e.profile()));
+        if (hasSlot) {
+            return;
+        }
+        List<String> pools = slotsFor(MemberRole.ARCHITECT).values().stream()
+                .map(Entry::profile)
+                .distinct()
+                .toList();
+        throw new IllegalArgumentException(
+                "no " + role.wireName() + " slot for profile '" + profile + "' — an architect's "
+                        + "identity IS the slot it is bound to, so there is nothing to bind this "
+                        + "session's identity to. fleet." + role.configKey() + " carries profiles: "
+                        + (pools.isEmpty() ? "(none configured)" : String.join(", ", pools))
+                        + "; add profile '" + profile + "' there, or spawn " + role.wireName()
+                        + " on one of those profiles instead");
     }
 
     /** Unbind a released terminal using the compare-safe registry operation. */

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

@@ -237,11 +237,13 @@ public final class CompletionResolver implements TurnListener {
         }
         String tail;
         String assistantBlock = null;
+        String rawScrape = null;
         int originalLength = 0;
         boolean clipped = false;
         boolean scrapeFailed = false;
         try {
-            assistantBlock = lastAssistantBlock(agents.read(target, SCRAPE_SOURCE));
+            rawScrape = agents.read(target, SCRAPE_SOURCE);
+            assistantBlock = lastAssistantBlock(rawScrape);
             originalLength = assistantBlock.strip().length();
             clipped = originalLength > MAX_SCRAPE_CHARS;
             tail = clip(assistantBlock);
@@ -256,6 +258,20 @@ public final class CompletionResolver implements TurnListener {
         // member, so a caller (including a lead deciding whether to delegate again) can tell a lost
         // turn from a real empty answer.
         if (scrapeFailed || tail.isEmpty()) {
+            // fleetd#211: lastAssistantBlock() found nothing usable — most often a pane with no ⏺
+            // marker at all, whose boundary scan then starts at the top of the raw screen and breaks
+            // immediately on the first line of TUI chrome (╭, │, ❯, …). Before giving up as a lost
+            // turn, run the same exhaustion/backend-error classification against the RAW scrape as a
+            // fallback, ONLY here. A pane that already yielded a usable block never reaches this
+            // branch, so the narrow (trimmed) match on the normal path below is completely unchanged
+            // — zero new false positives there. Every pane this fallback examines was already headed
+            // for the empty-scrape failure, so a wrong label here is strictly less bad than silently
+            // losing an exhaustion signal: the alternative outcome is already a failure, just one that
+            // never quarantines the credential. lastAssistantBlock stays the source of the reply
+            // TEXT everywhere else; only classification ever consults the raw scrape, and only here.
+            if (rawScrape != null && classifyRawScrapeFallback(target, turn, waiter, rawScrape)) {
+                return;
+            }
             fail(target, turn, emptyScrapeReason(target, scrapeFailed));
             return;
         }
@@ -313,6 +329,43 @@ public final class CompletionResolver implements TurnListener {
         }
     }
 
+    /**
+     * fleetd#211: the raw-scrape fallback classification, run only when {@link #lastAssistantBlock}
+     * found nothing usable (see the call site in {@link #resolve}). Mirrors the two classifications
+     * the normal path already applies to the trimmed assistant block — exhaustion first, then the
+     * narrow {@link #BACKEND_ERROR} pattern — against {@code raw} instead, and reports whether one of
+     * them handled the turn (resolved the waiter or failed it) so the caller skips the empty-scrape
+     * failure. Never runs on the normal (non-empty-block) path, and never touches the reply text.
+     */
+    private boolean classifyRawScrapeFallback(String target, InFlight turn,
+                                               CompletableFuture<Rendezvous.Resolution> waiter, String raw) {
+        Pattern exhausted = exhaustedPatterns.patternFor(target);
+        String matchedLine = exhausted == null ? null : firstMatchingLine(raw, exhausted);
+        if (matchedLine != null) {
+            String reason = "backend exhausted (usage limit): " + matchedLine;
+            if (rendezvous.resolveExhausted(waiter, reason)) {
+                inFlight.remove(target, turn);
+                log.warn("completion for {} classified BACKEND_EXHAUSTED from the raw scrape (no "
+                        + "usable assistant block; no fleet_reply): {}", target, reason);
+                // CB-578 stage B: only on the resolution that actually won the race — a late
+                // duplicate must never quarantine a credential twice for one refusal.
+                exhaustionSink.onExhausted(target, reason);
+            }
+            return true;
+        }
+        String backendError = firstMatchingLine(raw, BACKEND_ERROR);
+        if (backendError != null) {
+            // Carry the pane, not just the matched line — the same fleetd#164 rule the normal path
+            // above applies. Here it matters more, not less: the trimmed block was empty, so the raw
+            // scrape is the ONLY copy of whatever the member managed to say. Clipped to the same cap
+            // the normal path uses, since a raw screen has no boundary trimming to bound it.
+            fail(target, turn, "member " + target + " ended on a backend error: " + backendError
+                    + "\n--- pane tail ---\n" + clip(raw));
+            return true;
+        }
+        return false;
+    }
+
     /** Synchronous fail (the unit-testable core of {@link #onTurnFailed}). */
     void fail(String target, InFlight turn) {
         fail(target, turn, null);

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

@@ -278,18 +278,22 @@ public final class ClaudeCodeLauncher extends HerdrPeerLauncher {
 
     /**
      * Add the Claude-specific session-identity flags to {@code argv} and return the peer's OWN
-     * session id — the resume handle. A resume request passes the prior id via {@code -r} and
-     * returns that id; a fresh named session mints a new UUID, passes it via {@code --session-id},
-     * and returns the mint. The bridge's logical name rides along as {@code -n} when present. When
-     * <em>no</em> identity is requested (sessionName and resumeSessionId both blank) this adds
-     * nothing and returns {@code null}, keeping the legacy no-identity launch byte-identical.
+     * session id — the resume handle. A resume passes the prior id via {@code -r} and returns
+     * that id; every other spawn mints a new UUID, passes it via {@code --session-id}, and
+     * returns the mint. The bridge's logical name rides along as {@code -n} when present.
+     *
+     * <p>fleetd #214: the mint is unconditional. A plain {@code fleet_spawn} passes neither
+     * sessionName nor resumeSessionId, yet the member must still be resumable, and this id is
+     * the only resume handle a claude-code member has — unlike opencode, nothing resolves it
+     * after the launch. Checked against the real binary (claude 2.1.252): the flag is safe on
+     * every spawn. The binary takes only a valid UUID — it refuses any other value at argument
+     * parsing ("Invalid session ID. Must be a valid UUID.") — so the {@code UUID.randomUUID()}
+     * mint is required, not incidental. The only flag interaction the binary documents is with
+     * {@code -r} (both claim the session id), and the resume branch above never combines the two.
      */
     private static String applySessionIdentity(List<String> argv, String sessionName, String resumeSessionId) {
         boolean resuming = resumeSessionId != null && !resumeSessionId.isBlank();
         boolean named = sessionName != null && !sessionName.isBlank();
-        if (!resuming && !named) {
-            return null; // no identity requested — keep the legacy launch byte-identical
-        }
         if (named) {
             argv.add("-n");
             argv.add(sessionName);
@@ -320,11 +324,19 @@ public final class ClaudeCodeLauncher extends HerdrPeerLauncher {
      *
      * <p>CB-618: Claude Code refuses to start when BOTH {@code --append-system-prompt} and
      * {@code --append-system-prompt-file} are on the command line ("Cannot use both ... Please use
-     * only one"), so the two charters can never travel on separate flags. When both are present they
-     * are concatenated into the one file, role charter first and reply charter last — last is where
-     * the reply rule must sit, because it is the rule that must survive. When only the reply charter
-     * is present it keeps its proven inline {@code --append-system-prompt} delivery, which is also
-     * the only form that reaches a member with no repo checkout.
+     * only one"), so the two charters can never travel on separate flags. They are concatenated
+     * into the one file, role charter first and reply charter last — last is where the reply rule
+     * must sit, because it is the rule that must survive.
+     *
+     * <p>fleetd #220: a lone reply charter used to ride inline on {@code --append-system-prompt},
+     * which put ~800 bytes of prose on the command line herdr types into the pane. That line is
+     * capped at {@value HerdrPeerLauncher#PANE_COMMAND_BYTE_LIMIT} bytes by the pty itself, and
+     * everything past the cap is dropped with no error from any layer. The charter alone left about
+     * 50 bytes of headroom, so adding one flag ({@code --session-id}, fleetd #214) truncated the
+     * LAST argument instead — {@code --autocompact 250000} arrived as {@code --autocompact 25},
+     * claude rejected it, and every claude-code spawn died as an unexplained readiness timeout.
+     * The charter now always travels as a file, which takes the prose off the command line for
+     * good; {@link HerdrPeerLauncher#checkPaneCommandFits} is the backstop for whatever grows next.
      */
     private List<String> argvWithFleet(FleetConfig.Profile cfg, LaunchSpec spec) {
         String roleCharter = nonBlank(spec.roleCharter());
@@ -341,16 +353,13 @@ public final class ClaudeCodeLauncher extends HerdrPeerLauncher {
             argv.add("--mcp-config");
             argv.add(mcpConfigJson(cfg));
         }
-        // Combine the charters in order role -> reply, dropping any that are absent. When two
-        // or more survive they must ride one --append-system-prompt-file (CB-618 forbids the inline
-        // flag and the file flag together). A lone reply charter keeps its proven inline delivery.
+        // Combine the charters in order role -> reply, dropping any that are absent. They ride one
+        // --append-system-prompt-file (CB-618 forbids the inline flag and the file flag together),
+        // always — fleetd #220: charter prose on the command line overruns the pane's byte cap.
         List<String> charters = new java.util.ArrayList<>(2);
         if (roleCharter != null) charters.add(roleCharter);
         if (replyCharter != null) charters.add(replyCharter);
-        if (charters.size() == 1 && replyCharter != null && roleCharter == null) {
-            argv.add("--append-system-prompt");
-            argv.add(replyCharter);
-        } else if (!charters.isEmpty()) {
+        if (!charters.isEmpty()) {
             argv.add("--append-system-prompt-file");
             argv.add(writeCharterFile(String.join("\n\n", charters)).toString());
         }

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

@@ -682,6 +682,7 @@ 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();
@@ -697,6 +698,59 @@ public abstract class HerdrPeerLauncher implements PeerLauncher {
         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.");
+    }
+
+    /**
+     * 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}.
+     */
+    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;
@@ -860,20 +914,51 @@ public abstract class HerdrPeerLauncher implements PeerLauncher {
      */
     private void waitUntilInjectableOrThrow(String paneId) {
         long deadline = nowMillis.getAsLong() + spawnReadyTimeoutMs;
+        Object lastStatus = null;
         while (nowMillis.getAsLong() < deadline) {
-            if (agents.status(paneId).injectable()) {
+            var status = agents.status(paneId);
+            lastStatus = status;
+            if (status.injectable()) {
                 log.debug("peer pane={} reached injectable state", paneId);
                 return;
             }
             sleeper.run();
         }
-        log.warn("peer pane={} did not become injectable within {}ms — closing", paneId, spawnReadyTimeoutMs);
+        // fleetd #220: read the pane BEFORE stop() closes it. Without this the gate says only that
+        // it timed out, which is true of every cause — a backend that never launched, a binary that
+        // rejected an argument and exited, a trust prompt, a login shell that hung. The pane holds
+        // the one copy of that answer and it is destroyed a line later.
+        log.warn("peer pane={} did not become injectable within {}ms (last status {}) — closing. "
+                        + "Pane tail:\n{}",
+                paneId, spawnReadyTimeoutMs, lastStatus, readPaneQuietly(paneId));
         stop(paneId);
         throw new PeerUnreachableException(
                 "worker pane " + paneId + " did not reach injectable state within "
                         + spawnReadyTimeoutMs + "ms");
     }
 
+    /**
+     * fleetd #220: the pane's recent output, clipped, for the readiness-gate timeout log — or a
+     * short note when it cannot be read. Best-effort by construction: this runs on a path that is
+     * already failing, so it must never replace the real error with one of its own.
+     */
+    private String readPaneQuietly(String paneId) {
+        try {
+            String pane = agents.read(paneId, "recent");
+            if (pane == null || pane.isBlank()) {
+                return "<pane read returned nothing>";
+            }
+            return pane.length() <= SPAWN_FAILURE_PANE_CHARS
+                    ? pane
+                    : pane.substring(pane.length() - SPAWN_FAILURE_PANE_CHARS);
+        } catch (RuntimeException e) {
+            return "<pane could not be read: " + e.getMessage() + ">";
+        }
+    }
+
+    /** How much of a failed spawn's pane the timeout log carries. */
+    private static final int SPAWN_FAILURE_PANE_CHARS = 4000;
+
     /**
      * A concrete {@link PeerHandle} wrapping herdr agent coordinates, the profile that spawned it,
      * the session identity the launch resolved (CB-547a): the bridge's logical name and the peer's

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

@@ -183,6 +183,13 @@ public final class SessionManager implements TurnListener {
                                  String sessionName, String resumeSessionId) {
         MemberRole memberRole = (role == null) ? MemberRole.DEV : role;
         requireResumeCapability(profile, resumeSessionId);
+        // CB-619 / fleetd #123: an explicit profile bypasses placement (CompositePeerLauncher only
+        // constrains an UNQUALIFIED spawn to the role's pool), so it is the one path that can ask
+        // for a role with no slot to bind it to. Refuse before anything spawns. A blank profile is
+        // left to placement, which already restricts an unqualified spawn to the role's pool.
+        if (profile != null && !profile.isBlank()) {
+            memberLifecycle.requireSlotFor(memberRole, profile);
+        }
         if (wt == null) {
             // CB-557: the role must ride on the SpawnRequest, not stay a local. The launcher needs it
             // to pick the profile out of that role's pool and to label the tab; a role kept only on
@@ -198,11 +205,15 @@ public final class SessionManager implements TurnListener {
             String resolvedProfile = resolveProfile(handle, profile);
             String cwd = launcher.effectiveCwd(new SpawnRequest(resolvedProfile, requestedCwd, callerCwd));
             long now = nowNanos.getAsLong();
+            // CB-619: bind (or fail to bind) BEFORE the session is recorded, and store whatever role
+            // this call actually returns — never the requested memberRole — so the session's role,
+            // what GET /members and fleet_list report, is never a lie about what this terminal holds.
+            MemberRole actualRole = memberLifecycle.acquired(memberRole, resolvedProfile, handle.terminalId());
             MemberSession session = new MemberSession(
                     handle.id(),
                     handle.terminalId(),
                     resolvedProfile,
-                    memberRole,
+                    actualRole,
                     cwd,
                     ownerTerminal,
                     now,
@@ -215,7 +226,6 @@ public final class SessionManager implements TurnListener {
                     handle.agentSessionId());
             registry.put(handle.id(), session);
             handles.put(handle.id(), handle);
-            memberLifecycle.acquired(session.role(), session.profile(), session.terminalId());
             log.debug("acquired session id={} terminal={} profile={} owner={}",
                     handle.id(), handle.terminalId(), session.profile(), session.ownerTerminal());
             notifyAcquired(session.terminalId());
@@ -509,11 +519,14 @@ public final class SessionManager implements TurnListener {
         String resolvedProfile = resolveProfile(handle, profile);
         String cwd = launcher.effectiveCwd(new SpawnRequest(resolvedProfile, path, callerCwd));
         long now = nowNanos.getAsLong();
+        // CB-619: see the no-worktree path above — bind before recording, and store the returned
+        // actual role, so this session's role is never a lie about what it actually holds.
+        MemberRole actualRole = memberLifecycle.acquired(memberRole, resolvedProfile, handle.terminalId());
         MemberSession session = new MemberSession(
                 handle.id(),
                 handle.terminalId(),
                 resolvedProfile,
-                memberRole,
+                actualRole,
                 cwd,
                 ownerTerminal,
                 now,
@@ -526,7 +539,6 @@ public final class SessionManager implements TurnListener {
                 handle.agentSessionId());
         registry.put(handle.id(), session);
         handles.put(handle.id(), handle);
-        memberLifecycle.acquired(session.role(), session.profile(), session.terminalId());
         log.debug("acquired worktree session id={} terminal={} profile={} branch={} path={}",
                 handle.id(), handle.terminalId(), session.profile(), session.branch(), session.worktree());
         notifyAcquired(session.terminalId());

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

@@ -633,6 +633,115 @@ class CompletionResolverTest {
                 "the failure carries the rest of the pane, not only the matched line: " + reason);
     }
 
+    // --- fleetd#211: raw-scrape fallback classification when there is no usable assistant block ---
+
+    @Test
+    void anExhaustionLineWithNoMarkerAndLeadingChromeIsClassifiedFromTheRawScrapeAndNotifiesTheSink() {
+        // No ⏺ anywhere, and the first visible line is TUI chrome (╭). lastAssistantBlock's boundary
+        // scan starts at the top of the raw screen and breaks immediately, so the trimmed block is "".
+        // The fix: fall back to matching the RAW scrape so this doesn't get lost as an empty scrape.
+        String block = """
+                ╭──────────────────────────────────────╮
+                The usage limit has been reached. Try again later.
+                """;
+        FakeHerdr herdr = new FakeHerdr().readText(block);
+        Rendezvous rendezvous = new Rendezvous();
+        ExhaustedPatternLookup patterns = target -> Pattern.compile("usage limit has been reached");
+        java.util.List<String> notified = new java.util.ArrayList<>();
+        ExhaustionSink sink = (target, reason) -> notified.add(target + ": " + reason);
+        CompletionResolver resolver = new CompletionResolver(new AgentControl(herdr), rendezvous, patterns, sink);
+
+        var waiter = rendezvous.open("term_a");
+        resolver.resolve("term_a", new CompletionResolver.InFlight(waiter, null));
+
+        assertTrue(waiter.isDone(), "a raw-scrape match still resolves the blocked send");
+        assertEquals(Rendezvous.Kind.BACKEND_EXHAUSTED, waiter.getNow(null).kind(),
+                "classified from the raw scrape even though the trimmed block was empty");
+        assertEquals(1, notified.size(),
+                "the sink is the whole point of this ticket — it must be notified: " + notified);
+        assertTrue(notified.get(0).startsWith("term_a: "), "the sink is told which target exhausted");
+        assertTrue(notified.get(0).contains("The usage limit has been reached"),
+                "the sink is told the matched reason: " + notified.get(0));
+    }
+
+    @Test
+    void aBackendErrorLineWithNoMarkerAndLeadingChromeIsClassifiedFromTheRawScrape() {
+        String block = """
+                ╭──────────────────────────────────────╮
+                API Error: 400 invalid request body
+                """;
+        FakeHerdr herdr = new FakeHerdr().readText(block);
+        Rendezvous rendezvous = new Rendezvous();
+        CompletionResolver resolver = new CompletionResolver(new AgentControl(herdr), rendezvous, ExhaustedPatternLookup.none(), ExhaustionSink.none());
+
+        var waiter = rendezvous.open("term_a");
+        resolver.resolve("term_a", new CompletionResolver.InFlight(waiter, null));
+
+        assertTrue(waiter.isDone(), "a raw-scrape backend-error match still resolves the blocked send");
+        assertEquals(Rendezvous.Kind.FAILED, waiter.getNow(null).kind(),
+                "classified BACKEND_ERROR from the raw scrape even though the trimmed block was empty");
+        assertTrue(waiter.getNow(null).text().contains("API Error: 400 invalid request body"),
+                "the failure carries the matched line: " + waiter.getNow(null).text());
+        assertTrue(waiter.getNow(null).text().contains("--- pane tail ---"),
+                "fleetd#164: the failure must carry the pane, not only the matched line — with an "
+                        + "empty trimmed block the raw scrape is the only copy of what the member said: "
+                        + waiter.getNow(null).text());
+        assertTrue(waiter.getNow(null).text().contains("╭"),
+                "the carried pane is the raw scrape, chrome included: " + waiter.getNow(null).text());
+    }
+
+    @Test
+    void anOrdinaryPaneWithANormalAssistantBlockIsUnaffectedByTheRawScrapeFallback() {
+        // Pin: on a pane that already yields a usable block, the fallback branch is never reached —
+        // same outcome, same text, sink not called — even though the raw screen around the marker
+        // would itself match the configured exhausted pattern.
+        String block = "The usage limit has been reached, but this is a leading TUI line above the "
+                + "marker.\n⏺ complete report\n❯ ";
+        FakeHerdr herdr = new FakeHerdr().readText(block);
+        Rendezvous rendezvous = new Rendezvous();
+        ExhaustedPatternLookup patterns = target -> Pattern.compile("usage limit has been reached");
+        java.util.List<String> notified = new java.util.ArrayList<>();
+        ExhaustionSink sink = (target, reason) -> notified.add(target + ": " + reason);
+        CompletionResolver resolver = new CompletionResolver(new AgentControl(herdr), rendezvous, patterns, sink);
+
+        var waiter = rendezvous.open("term_a");
+        resolver.resolve("term_a", new CompletionResolver.InFlight(waiter, null));
+
+        assertTrue(waiter.isDone());
+        assertEquals(Rendezvous.Kind.COMPLETION, waiter.getNow(null).kind(),
+                "unchanged: a usable assistant block never reaches the raw-scrape fallback");
+        assertEquals("complete report", waiter.getNow(null).text(), "the reply text is unaffected");
+        assertTrue(notified.isEmpty(), "the fallback never runs, so the sink is never called");
+    }
+
+    @Test
+    void aGenuinelyEmptyScrapeStillFailsAsEmptyAndNeverNotifiesTheSink() {
+        // The false-positive pin: no exhaustion or backend-error text anywhere on the pane (just
+        // chrome, no marker) — the raw-scrape fallback must not manufacture a classification, and
+        // the sink must stay untouched.
+        String block = """
+                ╭──────────────────────────────────────╮
+                │ >                                    │
+                ╰──────────────────────────────────────╯
+                """;
+        FakeHerdr herdr = new FakeHerdr().readText(block);
+        Rendezvous rendezvous = new Rendezvous();
+        ExhaustedPatternLookup patterns = target -> Pattern.compile("usage limit has been reached");
+        java.util.List<String> notified = new java.util.ArrayList<>();
+        ExhaustionSink sink = (target, reason) -> notified.add(target + ": " + reason);
+        CompletionResolver resolver = new CompletionResolver(new AgentControl(herdr), rendezvous, patterns, sink);
+
+        var waiter = rendezvous.open("term_a");
+        resolver.resolve("term_a", new CompletionResolver.InFlight(waiter, null));
+
+        assertTrue(waiter.isDone(), "an empty scrape must still resolve the send, not hang");
+        assertEquals(Rendezvous.Kind.FAILED, waiter.getNow(null).kind(),
+                "no exhaustion or backend-error text anywhere ⇒ this stays the ordinary empty-scrape failure");
+        assertTrue(waiter.getNow(null).text().toLowerCase().contains("empty"),
+                "the failure still says the scrape was empty: " + waiter.getNow(null).text());
+        assertTrue(notified.isEmpty(), "a genuinely empty pane must never quarantine a credential");
+    }
+
     @Test
     void coverageIsOffWhenNoProfileHasAPatternConfigured() {
         assertEquals("off (no profile has an exhaustedPattern configured; profiles: [terra])",

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

@@ -1,10 +1,13 @@
 package dev.ltms.fleet.mcp;
 
+import dev.ltms.fleet.auth.CallerResolver;
+import dev.ltms.fleet.auth.MemberRegistry;
 import dev.ltms.fleet.auth.Principal;
 import dev.ltms.fleet.config.FleetConfig;
 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.inject.Injector;
 import dev.ltms.fleet.msg.MessageService;
@@ -1024,6 +1027,87 @@ class FleetMcpTest {
         assertTrue(textOf(res).contains("architect, dev, reviewer"), textOf(res));
     }
 
+    // ── CB-619 / fleetd #123: a spawn asking for a role its profile has no slot for must be
+    //    refused, never silently demoted with the roster still lying about it ───────────────────
+
+    /** Two profiles on one launcher, so a role's pool can name one and exclude the other. */
+    private static ClaudeCodeLauncher architectCapableLauncher(FakeHerdr h) {
+        FleetConfig.Profile opus = new FleetConfig.Profile(
+                "opus", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
+                "tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
+        FleetConfig.Profile sonnet = new FleetConfig.Profile(
+                "sonnet", "http://gx00.gw:8000", "coder", null, "FLEETD_WORKER_TOKEN", null,
+                "tab", "fleetd-workers", "worker: {profile} #{n}", null, null, null);
+        return new ClaudeCodeLauncher(new AgentControl(h), new WorkspaceControl(h),
+                new SubscriptionGuard(Set.of("gx00.gw")), Map.of("opus", opus, "sonnet", sonnet),
+                "sonnet", _ -> "tok");
+    }
+
+    /** {@code fleet.architects} carries only {@code opus} — {@code sonnet} has no matching slot. */
+    private static MemberRegistry architectRegistry() {
+        return new MemberRegistry(new FleetConfig.Fleet(Map.of(),
+                Map.of("opus", new FleetConfig.Slot("opus")), Map.of(), Map.of(), null));
+    }
+
+    /**
+     * The literal defect (fleetd #123): {@code role=architect, profile=sonnet}, where
+     * {@code fleet.architects} carries only {@code opus}. Drives the real path —
+     * {@link FleetMcp#spawn} calls {@link SessionManager#acquire}, which must refuse before ever
+     * reaching the real {@link ClaudeCodeLauncher} — never {@link MemberRegistry#bind} called
+     * directly, which would walk around the gate under test.
+     */
+    @Test
+    void spawnRefusesAnArchitectWithNoMatchingSlotAndNeverTouchesTheLauncher() {
+        FakeHerdr h = new FakeHerdr();
+        SessionManager sessions = new SessionManager(architectCapableLauncher(h));
+        sessions.setMemberLifecycle(architectRegistry());
+
+        McpSchema.CallToolResult res = FleetMcp.spawn(sessions, "sonnet", "architect",
+                null, null, null, null, null, null);
+
+        assertEquals(Boolean.TRUE, res.isError(), textOf(res));
+        String msg = textOf(res);
+        assertTrue(msg.contains("architect"), "names the role asked for: " + msg);
+        assertTrue(msg.contains("sonnet"), "names the profile: " + msg);
+        assertTrue(msg.contains("opus"), "names the pool that does carry the role: " + msg);
+        assertTrue(sessions.roster().isEmpty(), "a refused spawn must register no session");
+        assertTrue(h.calls.stream().noneMatch(c -> c.method().equals("agent.start")),
+                "a refused spawn must never reach the launcher — no process should ever start");
+    }
+
+    /**
+     * Positive control / parity check: when the profile DOES carry a slot, the spawn succeeds, and
+     * {@code GET /members} ({@link FleetMcp#listFleet}) and {@code fleet_whoami}
+     * ({@link FleetMcp#whoami}) — resolved through the SAME live {@link MemberRegistry} binding via
+     * a real {@link CallerResolver}, exactly as the daemon resolves a real MCP caller — must never
+     * disagree about this one live member's role.
+     */
+    @Test
+    void rosterAndWhoamiAgreeOnceTheArchitectSlotBinds() {
+        FakeHerdr h = new FakeHerdr();
+        // Pin the spawn onto the one pane FakeHerdr's canned pane.process_info maps to WORKER_PID,
+        // so a CallerResolver can resolve THIS session's own terminal, not a fixture double.
+        h.pinNextStarts(1, "term_a", "w2:p7");
+        SessionManager sessions = new SessionManager(architectCapableLauncher(h));
+        MemberRegistry members = architectRegistry();
+        sessions.setMemberLifecycle(members);
+
+        McpSchema.CallToolResult spawnRes = FleetMcp.spawn(sessions, "opus", "architect",
+                null, null, null, null, null, null);
+        assertNotEquals(Boolean.TRUE, spawnRes.isError(), textOf(spawnRes));
+        assertTrue(textOf(spawnRes).contains("\"role\":\"architect\""), textOf(spawnRes));
+
+        String roster = textOf(FleetMcp.listFleet(architectCapableLauncher(h), sessions, Map.of(), ""));
+        assertTrue(roster.contains("\"role\":\"architect\""), "GET /members: " + roster);
+
+        ConnectionIdentity identity = new ConnectionIdentity(new PaneLocator(h), _ -> FakeHerdr.WORKER_PID);
+        CallerResolver resolver = CallerResolver.withLeadsAndMembers(identity, false, null, Map::of, members);
+        Principal caller = resolver.resolve("127.0.0.1", 42, null);
+        assertTrue(caller.isArchitect(), "fleet_whoami's own resolver must agree the terminal is bound");
+        String whoami = textOf(FleetMcp.whoami(caller, sessions));
+        assertTrue(whoami.contains("\"role\":\"architect\""), "fleet_whoami: " + whoami);
+    }
+
     // ── CB-584: fleet_spawn accepts sessionName/resumeSessionId; roster shows agentSessionId ──
 
     @Test

fleetd/src/test/java/dev/ltms/fleet/member/ClaudeCodeLauncherTest.java

@@ -61,8 +61,75 @@ class ClaudeCodeLauncherTest {
         assertTrue(args.stream().noneMatch(a -> a.contains("\"bridge\"")),
                 "the mount is named fleet since CB-632 — a member addresses its tools as "
                         + "mcp__fleet__*, and CLAUDE.md's role-detection ladder names that prefix");
-        assertTrue(args.contains("--append-system-prompt"));
-        assertTrue(args.stream().anyMatch(a -> a.contains("fleet_reply")), "reply charter present");
+        // fleetd #220: the charter travels as a FILE, never inline — charter prose on the command
+        // line overruns the byte cap of the pane line herdr types it into.
+        assertTrue(args.contains("--append-system-prompt-file"));
+        assertFalse(args.contains("--append-system-prompt"),
+                "the inline flag would put ~800 bytes of prose on the pane command line");
+        String charterFile = args.get(args.indexOf("--append-system-prompt-file") + 1);
+        assertTrue(readFile(charterFile).contains("fleet_reply"), "reply charter present in the file");
+    }
+
+    /** Read a charter file the launcher wrote, failing the test rather than the build on an IO error. */
+    private static String readFile(String path) {
+        try {
+            return java.nio.file.Files.readString(java.nio.file.Path.of(path));
+        } catch (java.io.IOException e) {
+            throw new AssertionError("charter file " + path + " is not readable", e);
+        }
+    }
+
+    /**
+     * fleetd #220 regression. herdr TYPES the launch command into the pane, and the pty line buffer
+     * holds 1024 bytes — past that the tail is dropped with no error anywhere, so the backend exits
+     * on a mangled argument and the spawn dies as an unexplained readiness timeout. That is what
+     * happened when #214 added --session-id to a command already 978 bytes long: --autocompact
+     * 250000 arrived as --autocompact 25. This asserts the whole assembled command still fits, with
+     * the flags a real spawn carries (MCP mount, charter, model, autocompact, session id).
+     */
+    @Test
+    void theAssembledLaunchCommandFitsThePaneLineLimit() {
+        FakeHerdr herdr = new FakeHerdr();
+        // Shaped like the live sonnet profile, because the bug is a SUM: the charter alone fits,
+        // and so does every flag alone. Only model + autocompact + session id on top of the charter
+        // crossed the cap, which is why nothing caught it until a member failed to spawn.
+        FleetConfig.Profile cfg = new FleetConfig.Profile(
+                "sonnet", null, "claude-sonnet-5", null, null,
+                List.of("claude"), "tab", "fleetd-workers", "w #{n}", "http://127.0.0.1:8765/mcp",
+                null, null, null, null, null, null, null, null, true, null, null, null, null, null,
+                250000);
+        new ClaudeCodeLauncher(new AgentControl(herdr), new WorkspaceControl(herdr),
+                new SubscriptionGuard(Set.of()), Map.of(cfg.profile(), cfg), cfg.profile(), _ -> null)
+                .spawn();
+
+        List<String> args = spawnedArgs(herdr);
+        int bytes = "claude".length();
+        for (String arg : args) {
+            bytes += arg.getBytes(java.nio.charset.StandardCharsets.UTF_8).length + 3;
+        }
+        assertTrue(bytes <= HerdrPeerLauncher.PANE_COMMAND_BYTE_LIMIT,
+                "the launch command must fit the pane line: " + bytes + " bytes vs limit "
+                        + HerdrPeerLauncher.PANE_COMMAND_BYTE_LIMIT + " — args " + args);
+    }
+
+    /**
+     * fleetd #220: the guard refuses a command that cannot fit, instead of letting the pty drop the
+     * tail. The refusal must name the size and the argument to blame — a spawn that fails with
+     * "did not reach injectable state" tells the operator nothing, which is the whole reason this
+     * bug took a live pane scrape to find.
+     */
+    @Test
+    void anOverlongLaunchCommandIsRefusedWithTheSizeAndTheCulprit() {
+        FakeHerdr herdr = new FakeHerdr();
+        String huge = "x".repeat(1500);
+        ClaudeCodeLauncher launcher = service(herdr, List.of("claude", huge), null);
+
+        PeerUnreachableException refused = assertThrows(PeerUnreachableException.class, launcher::spawn);
+
+        assertTrue(refused.getMessage().contains("1024"), "names the limit: " + refused.getMessage());
+        assertTrue(refused.getMessage().contains("1500"), "names the culprit's size: " + refused.getMessage());
+        assertFalse(herdr.called("agent.start"),
+                "nothing may be started — a truncated command is worse than no spawn");
     }
 
     // CB-634: a profile with ideMcpUrl set mounts the IDE Index MCP as a second server and pins
@@ -262,10 +329,16 @@ class ClaudeCodeLauncherTest {
 
         Map<?, ?> start = (Map<?, ?>) herdr.lastCall("agent.start").params();
         assertEquals("claude", start.get("kind"), "herdr launches the canonical executable by kind");
-        // CB-533: the shared fixture pins model "coder", so the model flag is the whole args list.
         // What this test guards is that argv[0] is NOT repeated — herdr supplies it from `kind`.
-        assertEquals(List.of("--model", "coder"), start.get("args"),
-                "the configured executable is not repeated in args");
+        // fleetd #214: a plain spawn now also carries fleetd's own minted session id.
+        List<String> args = spawnedArgs(herdr);
+        assertFalse(args.contains("claude"), "the configured executable is not repeated in args");
+        int flag = args.indexOf("--session-id");
+        assertTrue(flag >= 0, "a plain spawn mints a session id: " + args);
+        assertDoesNotThrow(() -> UUID.fromString(args.get(flag + 1)), "the minted id is a valid UUID");
+        assertEquals(List.of("--model", "coder"),
+                List.of(args.get(args.size() - 2), args.get(args.size() - 1)),
+                "the CB-533 model flag still trails the launch flags: " + args);
     }
 
     @Test
@@ -278,7 +351,15 @@ class ClaudeCodeLauncherTest {
         assertFalse(args.contains("--append-system-prompt"), "no reply charter without mcpUrl");
         // CB-533: the model flag is independent of the MCP mount — pinning the model is not part of
         // "mount the bridge", so an unmounted worker still runs the model its profile names.
-        assertEquals(List.of("--verbose", "--model", "coder"), args,
+        // fleetd #214: a plain spawn now carries fleetd's own minted session id; drop the two
+        // mint elements when checking the rest of the argv.
+        int flag = args.indexOf("--session-id");
+        assertTrue(flag >= 0, "a plain spawn mints a session id even without a bridge mount: " + args);
+        assertDoesNotThrow(() -> UUID.fromString(args.get(flag + 1)), "the minted id is a valid UUID");
+        List<String> rest = new java.util.ArrayList<>(args);
+        rest.remove(flag + 1);
+        rest.remove(flag);
+        assertEquals(List.of("--verbose", "--model", "coder"), rest,
                 "the operator's own args are preserved, in order, ahead of the model flag");
     }
 
@@ -649,7 +730,7 @@ class ClaudeCodeLauncherTest {
         assertEquals("/work/proj", cwd, "effectiveCwd via SpawnRequest must match the three-arg resolution");
     }
 
-    // --- CB-547a: durable session identity (mint / resume / no-identity legacy) -----------------
+    // --- CB-547a / fleetd #214: durable session identity (always mint / resume) -----------------
 
     @Test
     void freshSpawnMintsASessionIdAndPassesTheName() {
@@ -684,18 +765,28 @@ class ClaudeCodeLauncherTest {
     }
 
     @Test
-    void noIdentitySpawnKeepsTheLegacyArgvAndCarriesNoSessionHandle() {
+    void plainSpawnMintsASessionIdSoEveryMemberIsResumable() {
+        // fleetd #214: a plain spawn passes no sessionName and no resumeSessionId, yet the member
+        // must still be resumable — the id is minted unconditionally, and it is the ONLY resume
+        // handle a claude-code member has (unlike opencode, nothing resolves it after the launch).
+        // The binary requires a valid UUID (checked against claude 2.1.252: a non-UUID is refused
+        // at argument parsing with "Invalid session ID. Must be a valid UUID.").
         FakeHerdr herdr = new FakeHerdr();
         ClaudeCodeLauncher svc = service(herdr, List.of("ccs", "ltms-local"), null);
 
         PeerHandle handle = svc.spawn(new SpawnRequest("ltms-local", null, null));
 
         List<String> args = spawnedArgs(herdr);
-        assertFalse(args.contains("--session-id"), "no identity → no --session-id");
-        assertFalse(args.contains("-n"), "no identity → no -n");
-        assertFalse(args.contains("-r"), "no identity → no -r");
-        assertNull(handle.agentSessionId(), "no identity → no resume handle");
-        assertNull(handle.sessionName(), "no identity → no logical name");
+        int flag = args.indexOf("--session-id");
+        assertTrue(flag >= 0 && flag + 1 < args.size(),
+                "--session-id is minted even when no identity is requested: " + args);
+        String minted = args.get(flag + 1);
+        assertDoesNotThrow(() -> UUID.fromString(minted), "--session-id is a valid UUID: " + minted);
+        assertEquals(minted, handle.agentSessionId(),
+                "the resume handle is the minted id, so every member is resumable from fleet_list");
+        assertFalse(args.contains("-n"), "no sessionName was requested → no -n");
+        assertFalse(args.contains("-r"), "no resume was requested → no -r");
+        assertNull(handle.sessionName(), "no sessionName was requested → no logical name");
     }
 
     @Test

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

@@ -4,6 +4,7 @@ import ch.qos.logback.classic.Level;
 import ch.qos.logback.classic.LoggerContext;
 import ch.qos.logback.classic.spi.ILoggingEvent;
 import ch.qos.logback.core.read.ListAppender;
+import dev.ltms.fleet.auth.MemberRegistry;
 import dev.ltms.fleet.config.FleetConfig;
 import dev.ltms.fleet.guard.SubscriptionGuard;
 import dev.ltms.fleet.herdr.AgentControl;
@@ -333,6 +334,57 @@ class SessionManagerTest {
         }
     }
 
+    /**
+     * CB-619 / fleetd #123: {@code requireSlotFor} closes the config-gap case (no slot at all
+     * carries the profile) before anything spawns, but a profile that DOES carry a slot can still
+     * lose the bind to a concurrent spawn racing for the same slot. This drives that residual case
+     * through the REAL path — {@link SessionManager#acquire} against the real {@link
+     * dev.ltms.fleet.member.ClaudeCodeLauncher} and {@link FakeHerdr} — never {@link
+     * dev.ltms.fleet.auth.MemberRegistry#bind} directly for the session under test (only the
+     * precondition uses it, to occupy the slot before the real spawn happens). The session that
+     * loses the race must be held as a plain {@code dev}, never left claiming {@code architect} in
+     * the roster, and the daemon log must say so at WARN.
+     */
+    @Test
+    void aSecondArchitectOnAnAlreadyBoundProfileIsHeldAsDevNotArchitectAndWarnsLoudly() {
+        LoggerContext ctx = (LoggerContext) LoggerFactory.getILoggerFactory();
+        ch.qos.logback.classic.Logger registryLog = (ch.qos.logback.classic.Logger)
+                LoggerFactory.getLogger(MemberRegistry.class);
+        ListAppender<ILoggingEvent> appender = new ListAppender<>();
+        appender.setContext(ctx);
+        appender.start();
+        registryLog.addAppender(appender);
+        registryLog.setLevel(Level.WARN);
+        try {
+            FakeHerdr herdr = new FakeHerdr();
+            SessionManager sessions = sessionManager(herdr);
+            MemberRegistry members = new MemberRegistry(
+                    new FleetConfig.Fleet(Map.of(), Map.of("opus", new FleetConfig.Slot("ltms-local")),
+                            Map.of(), Map.of(), null));
+            assertTrue(members.bind("architect:opus", "term_already_bound"),
+                    "precondition: occupy the sole architect slot before the real spawn under test");
+            sessions.setMemberLifecycle(members);
+
+            MemberSession session = sessions.acquire("ltms-local", MemberRole.ARCHITECT, null,
+                    "/caller", "term_primary", null);
+
+            assertEquals(MemberRole.DEV, session.role(),
+                    "the slot is taken, so this session must be held as a plain member, never a lie");
+            assertEquals("dev", SessionManager.rosterView(session, null).get("role"),
+                    "the roster must report what this session actually holds, not what it asked for");
+
+            String warn = appender.list.stream()
+                    .filter(e -> e.getLevel().equals(Level.WARN))
+                    .map(ILoggingEvent::getFormattedMessage)
+                    .findFirst()
+                    .orElse("no slot-exhaustion WARN logged");
+            assertTrue(warn.contains("ltms-local"), "the log names the profile: " + warn);
+            assertTrue(warn.contains(session.terminalId()), "the log names the terminal: " + warn);
+        } finally {
+            registryLog.detachAppender(appender);
+        }
+    }
+
     @Test
     void rosterReflectsAcquiredMinusReleased() {
         FakeHerdr herdr = new FakeHerdr();
@@ -887,15 +939,18 @@ class SessionManagerTest {
     }
 
     @Test
-    void acquireWithNeitherSessionFieldLeavesAgentSessionIdNull() {
+    void acquireWithNeitherSessionFieldStillMintsAnAgentSessionId() {
+        // fleetd #214: the claude-code launcher mints a session id for EVERY spawn, so a member is
+        // resumable even when the spawn asked for no session identity.
         FakeHerdr herdr = new FakeHerdr();
         SessionManager sessions = sessionManager(herdr);
 
         MemberSession s = sessions.acquire("ltms-local", null, null, null);
 
-        assertNull(s.agentSessionId(), "no identity requested — unchanged from before CB-584");
-        assertFalse(SessionManager.rosterView(s, null).containsKey("agentSessionId"),
-                "a null id is omitted from the roster, like charterSha256 for a receipt-less session");
+        assertNotNull(s.agentSessionId(),
+                "fleetd #214: a plain spawn mints a session id, so every member is resumable");
+        assertTrue(SessionManager.rosterView(s, null).containsKey("agentSessionId"),
+                "the minted id is in the roster, so fleet_list advertises every member's resume handle");
     }
 
     @Test