fleetd#323: close the reload-classifier drift with a reflection coverage test

ConfigRef.sameLaunchSettings' javadoc claimed it compares every component
the launcher reads at spawn. It missed ideProjectDir, ideOpenCommand and
autoCompactWindow, and changedDeferredKeys separately missed worktreeGroup
(baked into the same GitWorktrees as worktreeRoot, Fleetd.java:251). A
reload that changed only one of those keys reported "config reloaded" with
nothing deferred, and the running daemon kept the old value.

Fix the four instances, and add ConfigRefProfileCoverageTest: it enumerates
every FleetConfig.Profile record component by reflection, mutates each one
not in the new ConfigRef.LAUNCH_SETTINGS_EXCLUDED set on a base profile,
and asserts sameLaunchSettings actually notices — so a fifth missed field
fails the build by name instead of drifting silently. It also prints its
own denominator (26 components, 23 compared, 3 excluded) per the ticket's
requirement that a checker must be able to state what it checked.

Also add the `profile` field itself to the comparison (it was neither
compared nor excluded before this fix — the coverage test surfaced it).

Rewrote the sameLaunchSettings javadoc to describe what the coverage test
actually guarantees instead of repeating the unchecked claim.

.claude/skills/hunter/SKILL.md

@@ -0,0 +1,102 @@
+---
+name: hunter
+description: Defect-hunt procedure for a fleetd worker — sweep an assigned package for real bugs and report several ranked findings without fixing anything. Load this when the lead asks you to hunt or audit a scope rather than review one diff. Do NOT load `reviewer` for this; the two want different output.
+---
+
+# Hunter worker — procedure
+
+The turn contract (one `fleet_reply`, `fleet_ask` for the lead's decisions, honest reporting,
+never merge) is in **`CLAUDE.md` → Bridge communication → Worker** and already applies.
+
+**This skill is not `reviewer`.** `reviewer` judges one diff and reports the *single* most
+important issue in about 90 words. A hunt sweeps a whole package and reports *several* findings
+in a long structured form. Loading both gives you two contradictory output contracts, and the
+usual result is a worker that writes a good report into its terminal and ends the turn without
+sending it. Load exactly one.
+
+## 0. Read this before you read code: how the report gets home
+
+Your terminal reaches nobody. The lead sees **only** the text inside your `fleet_reply` call.
+
+A long report is exactly the case where this goes wrong, so plan for it:
+
+- **Write the report into the `fleet_reply` argument itself.** Do not compose it in your terminal
+  and then summarise it into the call.
+- If the report is long, **send it anyway** — one `fleet_reply` with everything.
+- If you end the turn without replying, the bridge scrapes your pane instead. That scrape carries
+  at most the last 4000 characters, and on a hunt it usually captures the tail of the lead's own
+  brief rather than your findings. The lead then has nothing and has to ask you again.
+
+## 1. Change nothing
+
+A hunt is read-only. Do not edit a production file, do not "quickly fix" what you find, and do
+not run a formatter. You may run the build and tests to *check* a claim, and you should say so
+when you did.
+
+## 2. Read the whole scope first
+
+Read every file in the assigned package before you judge any of it. A defect that a caller
+elsewhere in the same package makes unreachable is not a defect, and you cannot know that from
+one file.
+
+Stay inside the scope. If a defect there depends on a class outside it, read that class to
+confirm — but the defect itself must live in the scope you were given.
+
+## 3. The bar — this matters more than the count
+
+**Name the path into the bad state.** Say which caller, in which state, reaches it. A defect on
+paper is not a reachable defect. If you cannot name that path, keep the finding but mark it
+`unproven` and say exactly what you could not check. Do not drop it, and do not dress it up.
+
+**Say which direction the harm goes.** Data loss, privilege escalation and silent wrong answers
+are worth reporting even when the window is narrow. A finding whose worst outcome is a worse log
+line is not worth a block.
+
+Two workers once ran the same scope: the one that applied the direction-of-harm filter found ten
+real defects, the one that did not found none. Fewer findings the lead can act on beat many the
+lead has to triage.
+
+## 4. Shapes that have produced real merged fixes here
+
+Read for these first:
+
+1. **A one-way gate.** A guard added after an incident closes only the direction that incident
+   came from. Do not only ask what closes the gate — ask **which states still open it**.
+2. **A value read once, then used later to authorise something destructive**, after something
+   else has had a chance to change it.
+3. **A failure downgraded to a value that looks like a legitimate result** — `-1`, `null`, an
+   empty list, `false` — which a caller then trusts.
+4. **A lock held for one half of a read-modify-write and not the other**, or two collections
+   updated under different locks.
+5. **A comment or javadoc stating an invariant the code no longer keeps.** Comments are
+   load-bearing in this repo; a stale one has already caused a bug.
+
+## 5. What you cannot check, and must not claim you did
+
+- `fleetd/fleetd.yaml` is gitignored and **absent from your worktree**. You cannot read it. If a
+  finding depends on live configuration, name the key and say you could not check it.
+- `.mcp.json`, `opencode.json` and `.autoenv` in your worktree are neutralised stubs, not the
+  repo's real files.
+- The `wiki/` submodule pointer is months old. Do not cite it.
+
+Reporting a fact you took from the lead's brief as something you measured yourself is a false
+report, even when the fact is correct. Say where each fact came from.
+
+## 6. The report — what goes in `fleet_reply`
+
+One block per finding, most severe first:
+
+```
+FINDING N — <one line>
+file:line
+Path in:        <which caller, in which state, reaches this>
+Direction:      <data loss | escalation | silent wrong answer | outage | ...>
+Window/trigger: <when it actually happens>
+Confidence:     <confirmed by reading | unproven — say what you could not check>
+Why nothing else catches it: <the guard or test you checked, and why it misses>
+```
+
+End with one line naming every file you read, so the lead knows the denominator.
+
+**Nothing clears the bar?** Reply `NO FINDINGS`, name the files you read, and say what you ruled
+out. A clean sweep is a valid result; an invented defect is worse than none.

.claude/skills/reviewer/SKILL.md

@@ -10,6 +10,11 @@ never merge) is in **`CLAUDE.md` → Bridge communication → Worker** and alrea
 skill is only the *review procedure*: how to work the scope, and the exact shape of what you
 send back.
 
+**Wrong skill for a sweep.** This one reviews *one* diff or scope and reports the *single* most
+important issue. If the lead asked you to hunt or audit a whole package for several defects, load
+`hunter` instead and ignore this file — the two want different output, and following both is how a
+worker ends its turn with a good report that never gets sent.
+
 ## 1. Read the whole scope before you judge
 
 The delegation names your scope — a file, a diff, a PR, a function. **Read all of it first.**

CLAUDE.md

@@ -200,8 +200,13 @@ must obey belongs in the charter, not here.
   adapter, with a message naming the credential and the remaining seconds ("cooling off after
   repeated backend errors") — distinct wording from a quarantine refusal, so don't conflate the
   two when reading a spawn failure.
-- **Skills available to delegate:** `implementer` (worktree → commit → push → own PR) and
-  `reviewer` (scoped review → one structured finding). Name one in every delegation.
+- **Skills available to delegate:** `implementer` (worktree → commit → push → own PR),
+  `reviewer` (one diff → one structured finding) and `hunter` (sweep a package → several ranked
+  findings, change nothing). Name exactly one in every delegation. **`reviewer` and `hunter` are
+  not interchangeable** — `reviewer` caps the answer at one finding in about 90 words, so naming
+  it for a multi-finding sweep hands the worker two contradictory output contracts. That has
+  already cost three workers' turns: each wrote a good report to its terminal and ended the turn
+  with no `fleet_reply`, and the scrape returned the tail of the brief instead.
 - **Primary-side skills** (not delegation playbooks — a worker cannot use them):
   `port-to-opencode` (make an OpenCode session a participant in this workspace) and
   `fleets-status` (report every fleet that shares one LavinMQ instance).

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

@@ -236,7 +236,15 @@ public final class CallerResolver {
         // loopback-trust: same-host callers that are not workers are the primary. A non-loopback
         // caller is anonymous even here — and startup refuses that combination anyway
         // (FleetConfig.validateAuthExposure), so this is defence in depth, not the control.
-        return isLoopback(remoteAddr) ? Principal.primary(c.pid()) : Principal.anonymous();
+        //
+        // fleetd #317: "not a worker" must not be conflated with "identity unresolved". The real
+        // primary is a real process — its pid resolves (c.resolved()), it just owns no herdr pane.
+        // A caller whose peer-PID lookup failed (LsofPeerPidLookup's -1 sentinel — on any failure,
+        // silently including "lsof found no match") has no such pid, and PaneLocator's own javadoc
+        // already names what happens if that case is handed the primary role: a worker→primary
+        // escalation. So an unresolved caller is refused (ANONYMOUS — the same clean, already-tested
+        // "authenticated as nothing" outcome used everywhere else in this method), never promoted.
+        return isLoopback(remoteAddr) && c.resolved() ? Principal.primary(c.pid()) : Principal.anonymous();
     }
 
     private boolean presentedTokenMatches(String authorizationHeader) {

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

@@ -39,12 +39,18 @@ import java.util.function.Supplier;
  *       keeps the old value until a restart: {@code lifecycle:}, {@code leadHeartbeat:},
  *       {@code spawnReadyTimeoutMs} / {@code spawnReadyPollMs}, {@code quarantineCooldownSeconds}
  *       (CB-578 stage B — baked once into the {@code BackendQuarantine} built at startup),
- *       {@code guard:}, {@code worktreeRoot:}, adding or removing a profile (a new backend needs its own launcher,
+ *       {@code guard:}, {@code worktreeRoot:} and {@code worktreeGroup:} (both baked once into the
+ *       {@code GitWorktrees} built at {@code Fleetd.java:251} and never rebuilt — fleetd #323
+ *       instance 2 found {@code worktreeGroup} missing from this list and from
+ *       {@link #changedDeferredKeys}), adding or removing a profile (a new backend needs its own launcher,
  *       which is constructed once), <em>and an existing profile's launch settings</em> —
  *       {@code model}, {@code baseUrl}, {@code argv}, {@code env}, {@code mcpUrl},
  *       {@code exhaustedPattern} (CB-578 stage A — compiled once into {@code Fleetd.main}'s
  *       pattern map at startup), {@code errorPattern} (fleetd #201 Unit 5 — compiled once into
- *       {@code Fleetd.main}'s backend-error pattern map at startup, the same way), and the rest.
+ *       {@code Fleetd.main}'s backend-error pattern map at startup, the same way),
+ *       {@code ideProjectDir} / {@code ideOpenCommand} / {@code autoCompactWindow} (fleetd #323
+ *       instance 1 — all three are read at spawn off the same frozen profile map and were missing
+ *       from {@link #sameLaunchSettings}), and the rest of {@link #sameLaunchSettings}.
  *       {@code credentialId} (CB-578 stage B) is NOT on
  *       this list — it is read live off the config supplier at every quarantine check and
  *       exhaustion event, exactly like {@code weight} / {@code maxLoad}, so it is hot instead.
@@ -221,6 +227,13 @@ public final class ConfigRef implements Supplier<FleetConfig> {
         if (!Objects.equals(old.worktreeRoot(), fresh.worktreeRoot())) {
             changed.add("worktreeRoot");
         }
+        // Baked into the same GitWorktrees as worktreeRoot (Fleetd.java:251) and never rebuilt
+        // either — see the class doc. Missing this check was fleetd #323 instance 2: a reload
+        // that changed only worktreeGroup reported "config reloaded" with nothing deferred, and
+        // newly provisioned worktrees kept the old sharing behaviour.
+        if (!Objects.equals(old.worktreeGroup(), fresh.worktreeGroup())) {
+            changed.add("worktreeGroup");
+        }
         if (!Objects.equals(old.spawnReadyTimeoutMs(), fresh.spawnReadyTimeoutMs())
                 || !Objects.equals(old.spawnReadyPollMs(), fresh.spawnReadyPollMs())) {
             changed.add("spawnReady*");
@@ -265,14 +278,34 @@ public final class ConfigRef implements Supplier<FleetConfig> {
     }
 
     /**
-     * Whether two versions of a profile would launch a peer identically. Compares every component
-     * the launcher reads at spawn; {@code weight}, {@code maxLoad} and {@code credentialId} are
-     * excluded because those are read live (by the placement policy and, for credentialId, by
-     * {@code CompositePeerLauncher}/the CB-578 stage B exhaustion sink) and really do take effect on
-     * the next spawn.
+     * {@link FleetConfig.Profile} record components deliberately left out of
+     * {@link #sameLaunchSettings} because they are read <em>live</em>, not baked in at spawn — see
+     * the class doc's <em>Hot</em> bullet. {@code weight} and {@code maxLoad} are read live by the
+     * placement policy on every spawn; {@code credentialId} is read live by
+     * {@code CompositePeerLauncher} and the CB-578 stage B exhaustion sink. Nothing else is
+     * excluded — see {@code sameLaunchSettingsComparesEveryProfileComponentOrExcludesIt} in
+     * {@code ConfigRefProfileCoverageTest}, which enumerates every {@code Profile} record component
+     * by reflection and fails the build if one is neither compared below nor named here.
      */
-    private static boolean sameLaunchSettings(FleetConfig.Profile a, FleetConfig.Profile b) {
-        return Objects.equals(a.baseUrl(), b.baseUrl())
+    static final Set<String> LAUNCH_SETTINGS_EXCLUDED = Set.of("weight", "maxLoad", "credentialId");
+
+    /**
+     * Whether two versions of a profile would launch a peer identically.
+     *
+     * <p>This must compare every {@link FleetConfig.Profile} record component except the three in
+     * {@link #LAUNCH_SETTINGS_EXCLUDED}. That is not a claim this javadoc can make good on by
+     * itself — a javadoc saying "compares every component" is exactly what fleetd #323 found to be
+     * false for three fields (and a sibling method's field list, for a fourth). The actual
+     * guarantee comes from {@code ConfigRefProfileCoverageTest}: it enumerates every record
+     * component of {@code FleetConfig.Profile} by reflection, mutates each one not in
+     * {@code LAUNCH_SETTINGS_EXCLUDED} on a base profile, and asserts this method reports a
+     * difference — so a new component that is neither compared here nor added to
+     * {@code LAUNCH_SETTINGS_EXCLUDED} (with a reason) fails that test by name, rather than
+     * silently reporting "config reloaded" for a value the daemon never picked up.
+     */
+    static boolean sameLaunchSettings(FleetConfig.Profile a, FleetConfig.Profile b) {
+        return Objects.equals(a.profile(), b.profile())
+                && Objects.equals(a.baseUrl(), b.baseUrl())
                 && Objects.equals(a.model(), b.model())
                 && Objects.equals(a.configDir(), b.configDir())
                 && Objects.equals(a.tokenEnv(), b.tokenEnv())
@@ -298,6 +331,16 @@ public final class ConfigRef implements Supplier<FleetConfig> {
                 // fleetd #201 Unit 5: errorPattern is compiled once into Fleetd.main's backend-error
                 // pattern map at startup (see BackendErrorPatternLookup wiring), the same way
                 // exhaustedPattern is — a reload never re-reads it either.
-                && Objects.equals(a.errorPattern(), b.errorPattern());
+                && Objects.equals(a.errorPattern(), b.errorPattern())
+                // fleetd #323 instance 1: ideProjectDir and ideOpenCommand are read at spawn off the
+                // same frozen profile map as ideMcpUrl above (ClaudeCodeLauncher.java:267/269,
+                // OpenCodeLauncher.java:474/480/486) and were missing from this comparison.
+                && Objects.equals(a.ideProjectDir(), b.ideProjectDir())
+                && Objects.equals(a.ideOpenCommand(), b.ideOpenCommand())
+                // fleetd #323 instance 1: autoCompactWindow is read at spawn the same way
+                // (ClaudeCodeLauncher.java:926, OpenCodeLauncher.java:650). Comparing it here only
+                // makes the reload REPORT that a restart is needed — it deliberately does not make
+                // autoCompactWindow take effect live, which is a separate, larger change.
+                && Objects.equals(a.autoCompactWindow(), b.autoCompactWindow());
     }
 }

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

@@ -36,6 +36,25 @@ public final class ConnectionIdentity {
      * primary / an off-host client) and its {@code pid} (or {@code -1} if not resolvable).
      */
     public record Caller(String terminal, long pid) {
+
+        /**
+         * Whether the OS peer-PID lookup actually succeeded — {@code false} means {@code pid} is
+         * the {@code -1} sentinel, not a real process id, so this caller's identity could not be
+         * established at all. That is a different fact from a real pid that simply owns no worker
+         * pane (the primary's own connection): the primary is {@code resolved()} and has a
+         * {@code null terminal}; an unresolvable caller is {@code !resolved()} and also has a
+         * {@code null terminal}. The two look identical through {@link #terminal} alone, which is
+         * exactly how fleetd #317 happened — a failed {@code lsof} lookup and a genuine primary both
+         * fell through to {@code Principal.primary(...)}.
+         *
+         * <p>Centralised here, next to the sentinel it tests, for the same reason
+         * {@link ConnectionIdentity#isLoopback} is centralised rather than left for each caller to
+         * reimplement: a raw {@code pid > 0} check duplicated at every call site is precisely the
+         * "one rule, two copies" shape that let #305 drift.
+         */
+        public boolean resolved() {
+            return pid > 0;
+        }
     }
 
     /** Resolve the caller's terminal and PID from one peer-PID lookup. */
@@ -71,13 +90,27 @@ public final class ConnectionIdentity {
      * {@code 127.0.0.1:8765} with a source address of {@code 127.0.0.2} — measured on the Linux
      * fleet host, where binding that source succeeds.
      *
-     * <p><strong>Being strict here does not make the daemon safer; it makes it unsafe.</strong>
-     * That reads backwards, so it is worth stating plainly. This predicate does not decide whether
-     * a caller is trusted — it decides whether the caller's identity is <em>resolved at all</em>.
-     * Returning false means {@link #resolve} answers "no terminal", and downstream a caller with no
-     * terminal is treated as the primary under loopback-trust. So every address excluded here is an
-     * address on which a worker silently becomes the lead. Widening a check normally weakens it;
-     * widening this one is what closes the hole.
+     * <p><strong>What excluding an address costs, stated as it is today.</strong> This paragraph
+     * used to say that narrowing this range turned a worker into the lead, and that widening the
+     * check was what closed the hole. That was true only while there were <em>two</em> definitions
+     * that disagreed: {@code ConnectionIdentity} skipped the identity lookup for {@code 127.0.0.2}
+     * while {@code CallerResolver} read the same address as loopback and granted the primary role.
+     * #305 removed the second copy, and with one shared definition the old sentence no longer holds.
+     *
+     * <p>Measured on 2026-09-04 by narrowing this method back to exactly {@code 127.0.0.1} and
+     * running {@code CallerResolverTest} and {@code ConnectionIdentityTest}: a caller from
+     * {@code 127.0.0.2} then resolves to {@code ANONYMOUS}, not {@code PRIMARY} — for a worker
+     * ({@code aWorkerOnAnyLoopbackSourceAddressIsStillAWorkerNotThePrimary}) and for a non-worker
+     * ({@code aNonWorkerOnAnyLoopbackSourceAddressIsStillThePrimary}) alike. Excluding an address
+     * now <em>refuses</em> its caller; it does not promote one.
+     *
+     * <p>So keep the whole range, but for the plain reason: a genuine worker or primary that
+     * connects from {@code 127.0.0.2} must be identifiable at all, and narrowing this predicate
+     * locks it out. That is an outage, and an outage is the direction to fail in — which is exactly
+     * why the range must not be narrowed casually and also why doing so is no longer a security
+     * hole. This predicate still does not decide whether a caller is trusted; it decides whether the
+     * caller's identity is <em>resolved at all</em>. What makes an unresolved caller safe is
+     * {@link Caller#resolved()} (#317), not this method.
      */
     public static boolean isLoopback(String addr) {
         if (addr == null) {

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

@@ -41,6 +41,14 @@ public final class LsofPeerPidLookup implements PeerPidLookup {
             if (!p.waitFor(2, TimeUnit.SECONDS)) {
                 p.destroyForcibly();
             }
+            if (found < 0) {
+                // fleetd #317: this is the silent path — lsof ran clean and simply reported no
+                // matching process (e.g. queried before the OS socket table settles). Previously
+                // this logged nothing at all, which is exactly why the escalation went unnoticed;
+                // the exception path below already logs. A caller now refused because of this is
+                // still refused (never promoted) — this line only makes the refusal diagnosable.
+                log.debug("lsof peer-pid lookup for port {} found no matching process", port);
+            }
             return found;
         } catch (Exception e) {
             log.debug("lsof peer-pid lookup for port {} failed: {}", port, e.getMessage());

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

@@ -103,6 +103,54 @@ class CallerResolverTest {
         assertEquals(Role.PRIMARY, p.role(), "the historical behaviour, now an explicit choice");
     }
 
+    // ── fleetd #317: an unresolvable caller must never be promoted to the primary ──────────────────
+    // #305 closed the trigger where a resolved pid matched no pane *and* had no ancestry walk to
+    // save it. This is the other trigger PaneLocator's javadoc names: the pid never resolves at
+    // all — LsofPeerPidLookup returns -1 on any failure, including (silently) "lsof found no
+    // match" — so there is no candidate pid for an ancestry walk to even attempt.
+
+    /**
+     * The failing-without-the-fix case. Before #317's fix, {@code c.terminal() == null} was the
+     * only test in the loopback-trust fallback, and an unresolved pid produces exactly that same
+     * {@code null} terminal as a genuine primary — so this caller was handed
+     * {@code Principal.primary(...)}, a real worker's failed lookup becoming indistinguishable from
+     * the lead.
+     */
+    @Test
+    void aFailedPeerPidLookupIsRefusedNotPromotedToPrimary() {
+        ConnectionIdentity unresolved = new ConnectionIdentity(new PaneLocator(herdr), _ -> -1);
+        Principal p = new CallerResolver(unresolved).resolve("127.0.0.1", 55555, null);
+
+        assertEquals(Role.ANONYMOUS, p.role(),
+                "an unresolvable caller must never be silently promoted to the primary");
+    }
+
+    /**
+     * The companion invariant #317 must not break: a caller whose lookup genuinely succeeded, and
+     * who simply owns no herdr pane — the real primary's own connection — is still the primary.
+     * This is {@link #loopbackTrustTreatsANonWorkerLoopbackCallerAsThePrimary} pinned again here,
+     * named for #317 and placed next to the test it must be distinguished from: same {@code null}
+     * terminal, opposite verdict, because {@code Caller.resolved()} tells them apart.
+     */
+    @Test
+    void aRealPidThatOwnsNoPaneIsStillThePrimaryNotRefused() {
+        Principal p = new CallerResolver(nonWorkerIdentity()).resolve("127.0.0.1", 55555, null);
+
+        assertEquals(Role.PRIMARY, p.role());
+    }
+
+    /** #317 point 4: token mode never consults {@code c.pid()}, so a failed lookup must not change it. */
+    @Test
+    void tokenModeIsUndisturbedByAnUnresolvedLookup() {
+        ConnectionIdentity unresolved = new ConnectionIdentity(new PaneLocator(herdr), _ -> -1);
+        CallerResolver r = new CallerResolver(unresolved, true, "s3cret");
+
+        assertEquals(Role.ANONYMOUS, r.resolve("127.0.0.1", 55555, null).role(),
+                "no credential is still just ANONYMOUS, as before #317 — unchanged by the lookup failing");
+        assertEquals(Role.PRIMARY, r.resolve("127.0.0.1", 55555, "Bearer s3cret").role(),
+                "a valid token still authenticates the primary even though the peer-pid lookup failed");
+    }
+
     @Test
     void tokenModeRefusesANonWorkerCallerThatPresentsNoToken() {
         Principal p = new CallerResolver(nonWorkerIdentity(), true, "s3cret")

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

@@ -0,0 +1,187 @@
+package dev.ltms.fleet.config;
+
+import org.junit.jupiter.api.Test;
+
+import java.lang.reflect.Constructor;
+import java.lang.reflect.RecordComponent;
+import java.util.Arrays;
+import java.util.LinkedHashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+import java.util.TreeSet;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+/**
+ * fleetd #323: {@code ConfigRef.sameLaunchSettings} javadoc used to claim it "compares every
+ * component the launcher reads at spawn". It did not — {@code ideProjectDir}, {@code
+ * ideOpenCommand} and {@code autoCompactWindow} were all baked in at daemon startup (see
+ * {@code ClaudeCodeLauncher}/{@code OpenCodeLauncher}) and missing from the comparison, so a reload
+ * that changed only one of them reported "config reloaded" and the running daemon kept the old
+ * value.
+ *
+ * <p>This class is the mechanism the issue asked for: it enumerates every record component of
+ * {@link FleetConfig.Profile} by reflection and proves — by actually mutating a base profile one
+ * field at a time and calling the real method — that each component is either compared by
+ * {@link ConfigRef#sameLaunchSettings} or named in {@link ConfigRef#LAUNCH_SETTINGS_EXCLUDED} with
+ * a reason. A new profile field that is neither fails this test by name, not a hand-maintained list
+ * going stale.
+ */
+class ConfigRefProfileCoverageTest {
+
+    private static final RecordComponent[] COMPONENTS = FleetConfig.Profile.class.getRecordComponents();
+
+    /**
+     * One valid, non-blank value per record component — "the a value". None of these trip any
+     * defaulting/normalization in {@code Profile}'s compact constructor (see {@code
+     * FleetConfig.java}), so what goes in is what {@code sameLaunchSettings} sees back out.
+     */
+    private static final Map<String, Object> BASE = baseValues();
+
+    /** The same shape, each value distinct from {@link #BASE} — "the b value". */
+    private static final Map<String, Object> ALT = altValues();
+
+    private static Map<String, Object> baseValues() {
+        Map<String, Object> v = new LinkedHashMap<>();
+        v.put("profile", "sonnet");
+        v.put("baseUrl", "http://gx00.gw:8000");
+        v.put("model", "sonnet");
+        v.put("configDir", "/config/a");
+        v.put("tokenEnv", "TOKEN_A");
+        v.put("argv", List.of("claude", "--flag-a"));
+        v.put("placement", "tab");
+        v.put("workspace", "workspace-a");
+        v.put("tabLabel", "label-a");
+        v.put("mcpUrl", "http://mcp-a");
+        v.put("cwd", "/cwd/a");
+        v.put("parityOverlay", List.of(".env", ".env.a"));
+        v.put("gitTokenEnv", "GIT_TOKEN_A");
+        v.put("gitHostEnv", "GITEA_HOST_A");
+        v.put("kind", "claude-code");
+        v.put("env", Map.of("K", "A"));
+        v.put("weight", 1.0f);
+        v.put("maxLoad", 5);
+        v.put("subscription", Boolean.TRUE);
+        v.put("exhaustedPattern", "usage limit a");
+        v.put("credentialId", "cred-a");
+        v.put("ideMcpUrl", "http://ide-mcp-a");
+        v.put("ideProjectDir", "modules/a");
+        v.put("ideOpenCommand", "open-cmd-a {dir}");
+        v.put("autoCompactWindow", 150000);
+        v.put("errorPattern", "error a");
+        assertNamesMatchComponents(v);
+        return v;
+    }
+
+    private static Map<String, Object> altValues() {
+        Map<String, Object> v = new LinkedHashMap<>();
+        v.put("profile", "sonnet-b");
+        v.put("baseUrl", "http://gx01.gw:8000");
+        v.put("model", "haiku");
+        v.put("configDir", "/config/b");
+        v.put("tokenEnv", "TOKEN_B");
+        v.put("argv", List.of("claude", "--flag-b"));
+        v.put("placement", "weighted");
+        v.put("workspace", "workspace-b");
+        v.put("tabLabel", "label-b");
+        v.put("mcpUrl", "http://mcp-b");
+        v.put("cwd", "/cwd/b");
+        v.put("parityOverlay", List.of(".env", ".env.b"));
+        v.put("gitTokenEnv", "GIT_TOKEN_B");
+        v.put("gitHostEnv", "GITEA_HOST_B");
+        v.put("kind", "opencode");
+        v.put("env", Map.of("K", "B"));
+        v.put("weight", 2.0f);
+        v.put("maxLoad", 9);
+        v.put("subscription", Boolean.FALSE);
+        v.put("exhaustedPattern", "usage limit b");
+        v.put("credentialId", "cred-b");
+        v.put("ideMcpUrl", "http://ide-mcp-b");
+        v.put("ideProjectDir", "modules/b");
+        v.put("ideOpenCommand", "open-cmd-b {dir}");
+        v.put("autoCompactWindow", 250000);
+        v.put("errorPattern", "error b");
+        assertNamesMatchComponents(v);
+        return v;
+    }
+
+    private static void assertNamesMatchComponents(Map<String, Object> values) {
+        Set<String> componentNames = new TreeSet<>();
+        for (RecordComponent rc : COMPONENTS) {
+            componentNames.add(rc.getName());
+        }
+        assertEquals(componentNames, new TreeSet<>(values.keySet()),
+                "this test's value map has drifted from FleetConfig.Profile's actual components — "
+                        + "update BASE/ALT alongside the record");
+    }
+
+    private static FleetConfig.Profile profileOf(Map<String, Object> values) throws ReflectiveOperationException {
+        Class<?>[] types = Arrays.stream(COMPONENTS).map(RecordComponent::getType).toArray(Class<?>[]::new);
+        Object[] args = Arrays.stream(COMPONENTS)
+                .map(rc -> values.get(rc.getName()))
+                .toArray();
+        Constructor<FleetConfig.Profile> ctor = FleetConfig.Profile.class.getDeclaredConstructor(types);
+        return ctor.newInstance(args);
+    }
+
+    /** {@code BASE} with exactly one named component swapped for its {@code ALT} value. */
+    private static FleetConfig.Profile mutate(String componentName) throws ReflectiveOperationException {
+        Map<String, Object> values = new LinkedHashMap<>(BASE);
+        values.put(componentName, ALT.get(componentName));
+        return profileOf(values);
+    }
+
+    /**
+     * The mechanism fleetd #323 asked for: enumerate {@link FleetConfig.Profile}'s record
+     * components, mutate each non-excluded one, and prove {@code sameLaunchSettings} actually
+     * notices — not just that some hand-maintained list claims it does. Prints the denominator
+     * (total / compared / excluded) the issue required: a checker that cannot state its own
+     * denominator is the failure this repo keeps hitting.
+     */
+    @Test
+    void sameLaunchSettingsComparesEveryProfileComponentOrExcludesIt() throws ReflectiveOperationException {
+        int total = COMPONENTS.length;
+        Set<String> excluded = ConfigRef.LAUNCH_SETTINGS_EXCLUDED;
+
+        Set<String> allNames = new TreeSet<>();
+        for (RecordComponent rc : COMPONENTS) {
+            allNames.add(rc.getName());
+        }
+        assertTrue(allNames.containsAll(excluded),
+                "ConfigRef.LAUNCH_SETTINGS_EXCLUDED names a component that does not exist on "
+                        + "FleetConfig.Profile — check for a typo: " + excluded);
+
+        FleetConfig.Profile base = profileOf(BASE);
+        List<String> uncovered = new java.util.ArrayList<>();
+        int compared = 0;
+        for (RecordComponent rc : COMPONENTS) {
+            String name = rc.getName();
+            if (excluded.contains(name)) {
+                continue;
+            }
+            FleetConfig.Profile mutated = mutate(name);
+            if (ConfigRef.sameLaunchSettings(base, mutated)) {
+                uncovered.add(name);
+            } else {
+                compared++;
+            }
+        }
+
+        System.out.printf(
+                "ConfigRef.sameLaunchSettings coverage — %d Profile components total, %d compared, "
+                        + "%d excluded (%s)%n",
+                total, compared, excluded.size(), excluded);
+
+        assertEquals(List.of(), uncovered,
+                "these FleetConfig.Profile components changed but ConfigRef.sameLaunchSettings "
+                        + "reported no difference — add each one to the comparison (it is read at "
+                        + "spawn and baked in until a restart) or to ConfigRef.LAUNCH_SETTINGS_EXCLUDED "
+                        + "with a reason it is genuinely read live: " + uncovered);
+        assertEquals(total, compared + excluded.size(),
+                "every FleetConfig.Profile record component must be either compared or excluded — "
+                        + total + " components, " + compared + " compared, " + excluded.size()
+                        + " excluded");
+    }
+}

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

@@ -434,6 +434,78 @@ class ConfigRefTest {
         assertEquals("provider 5xx", ref.get().profiles().get("sonnet").errorPattern());
     }
 
+    /**
+     * fleetd #323 instance 1: {@code ideProjectDir} is read at spawn off the frozen profile map
+     * (see {@code ClaudeCodeLauncher}/{@code OpenCodeLauncher}) exactly like {@code model}, but was
+     * missing from {@code sameLaunchSettings} — a reload changing only this field used to report a
+     * bare "config reloaded" and the running daemon kept launching with the old value.
+     */
+    @Test
+    void changingAProfilesIdeProjectDirIsReportedAsDeferred(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("fleetd.yaml");
+        Files.writeString(f, """
+                bind:
+                  host: 127.0.0.1
+                  port: 8765
+                herdrSocket: ~/.config/herdr/herdr.sock
+                profiles:
+                  sonnet:
+                    baseUrl: http://gx00.gw:8000
+                    model: sonnet
+                    ideProjectDir: fleetd
+                guard:
+                  offSubscriptionHosts:
+                    - gx00.gw
+                """);
+        ConfigRef ref = refFor(f);
+
+        Files.writeString(f, """
+                bind:
+                  host: 127.0.0.1
+                  port: 8765
+                herdrSocket: ~/.config/herdr/herdr.sock
+                profiles:
+                  sonnet:
+                    baseUrl: http://gx00.gw:8000
+                    model: sonnet
+                    ideProjectDir: fleetd-renamed
+                guard:
+                  offSubscriptionHosts:
+                    - gx00.gw
+                """);
+        ConfigRef.Outcome out = ref.reload();
+
+        assertTrue(out.applied());
+        assertEquals(1, out.deferred().size(), out.deferred().toString());
+        assertTrue(out.deferred().getFirst().contains("sonnet"), out.deferred().toString());
+        assertTrue(out.deferred().getFirst().contains("launch settings"), out.deferred().toString());
+        // The snapshot still carries the new value — a restart is what makes it take effect.
+        assertEquals("fleetd-renamed", ref.get().profiles().get("sonnet").ideProjectDir());
+    }
+
+    /**
+     * fleetd #323 instance 2: {@code worktreeGroup} is baked into the same {@code GitWorktrees}
+     * as {@code worktreeRoot} (Fleetd.java:251) and never rebuilt, but only {@code worktreeRoot}
+     * was on {@code changedDeferredKeys} — a reload changing only the group reported a bare
+     * "config reloaded" and newly provisioned worktrees kept the old sharing behaviour.
+     */
+    @Test
+    void changingWorktreeGroupIsReportedAsDeferred(@TempDir Path dir) throws Exception {
+        Path f = dir.resolve("fleetd.yaml");
+        Files.writeString(f, yaml("worktreeGroup: devgroup\n"));
+        ConfigRef ref = refFor(f);
+
+        Files.writeString(f, yaml("worktreeGroup: devgroup2\n"));
+        ConfigRef.Outcome out = ref.reload();
+
+        assertTrue(out.applied());
+        assertEquals(java.util.List.of("worktreeGroup"), out.deferred());
+        assertTrue(out.summary().contains("needs a restart") || out.summary().contains("need a restart"),
+                out.summary());
+        // The snapshot still carries the new value — a restart is what makes it take effect.
+        assertEquals("devgroup2", ref.get().worktreeGroup());
+    }
+
     @Test
     void aFixedRefHasNoFileAndRefusesToReload() {
         FleetConfig cfg = new FleetConfig(null, null, null, null, null, null,

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

@@ -43,6 +43,22 @@ class ConnectionIdentityTest {
         assertNull(with(_ -> 999_999).callerTerminal("127.0.0.1", 55555));
     }
 
+    @Test
+    void callerIsUnresolvedWhenThePeerPidLookupFails() {
+        // fleetd #317: LsofPeerPidLookup returns -1 on any failure — a fork error, or (silently)
+        // simply no matching lsof line. Caller.resolved() is the one place that sentinel is tested.
+        ConnectionIdentity.Caller c = with(_ -> -1).resolve("127.0.0.1", 55555);
+        assertFalse(c.resolved(), "a -1 pid means the lookup failed, not that this pid owns no pane");
+    }
+
+    @Test
+    void callerIsResolvedWhenThePidIsRealEvenThoughItOwnsNoPane() {
+        // The primary's own connection: a real, lsof-found pid that just isn't a worker pane. This
+        // must read as "resolved" — the distinction #317 turns on.
+        ConnectionIdentity.Caller c = with(_ -> 999_999).resolve("127.0.0.1", 55555);
+        assertTrue(c.resolved());
+    }
+
     @Test
     void resolvesTheCallersPidAndCwd() {
         // CB-112: the primary maps to no pane, but its PID and cwd are still readable.