@ai-dossier/sched 0.70.1-next.732.1 → 0.70.2-next.735.1

package/README.md

@@ -202,6 +202,31 @@ where every mechanical supervision decision is code, not remembered prose:
    tick; a runnable unit never waits while a slot is idle (pinned by a regression test).
    Refill was always synchronous; what previously had no journal trace was the release
    itself — see `slot-released` below (#525).
+   **Auto-upgrade and child supervision (#1244):** `sched start --auto-upgrade`
+   (including `--once`) defers installation and process replacement while any
+   recorded slot agent is alive, or this engine/its tick worker still owns a child
+   awaiting its exit callback. This includes parallel members, tail, fix and report
+   agents, even after their slot is released. Upgrades retry at later tick boundaries
+   once those agents drain; the old engine never writes after installing the update.
+   On Linux a `/proc/<pid>/stat` state of `Z` is dead for reconciliation, not an
+   unkillable agent. Owned children are reaped by libuv when their event loop runs;
+   detecting `Z` alone does not reap a child adopted from an older engine image.
+   If Linux reports inherited zombies after all live agents and known child
+   callbacks drain (including after a tick-worker crash), the engine exits with
+   code 75 instead of re-executing: the OS reaper collects the children and the
+   service/watchdog starts a fresh process on retained state. An unsupervised
+   engine must be started again by its operator. A deliberate upgrade deferral
+   is journaled as `engine-auto-upgrade-deferred` with recorded slot/PID and child
+   information; unchanged deferrals are deduplicated, and drain is recorded once.
+
+   To recover an existing `agent-unkillable` batch without abandoning its work:
+   `ai-dossier sched resume --batch <id> --project <slug>`. It rechecks every
+   recorded batch/member agent's PID and start identity under the state lock and
+   refuses while any is genuinely alive. Once all are dead (including zombies),
+   it resumes the retained executing position. Members, landed commits, worktrees,
+   branches and dispatch profile stay intact; the next tick finishes the original
+   member decision. This does not toggle the global pause state. Avoid `abandon`
+   when the goal is to keep the batch.
 6. **Journal (AC6)** — every event (assigned, spawned, exit-detected, external-advance,
    progress, stalled, redispatched, fence-written, fence-failed, unit-failed,
    dependents-blocked, slot-released, suspect-dispatch, dispatch-unhealthy, dispatch-failure,
@@ -403,8 +428,8 @@ ai-dossier sched service uninstall
   node / entry-point / repo paths that are missing, relative, or under a temp or `worktrees/`
   directory, and warns that an nvm node path must be re-pinned (re-run `install`) after a node
   upgrade. It runs `sched start --auto-upgrade` (opt out with `--no-auto-upgrade`); after a
-  completed tick, a stale engine installs the latest CLI and re-execs the same command before
-  another state write. Detached agents keep running and the new engine reattaches by pid. The
+   completed tick and after all recorded agents and owned children drain, a stale engine
+   installs the latest CLI and re-execs the same command before another state write. The
   new version migrates state on load; older engines reject newer schemas. If the Node runtime
   cannot replace the process directly, exit code 75 asks systemd/the watchdog to restart the
   updated entry point.

package/dist/batch-dispatch.d.ts

@@ -165,6 +165,16 @@ export interface MemberFailure {
      */
     kind?: MemberExitKind;
 }
+/** #1244: this block originates only in executing's member stop ladder.
+ * Keep slots and member runs so normal reconciliation finishes that exact
+ * decision, rather than redispatching members or discarding landed work.
+ * Liveness and transition share the state lock with dispatch/stop.
+ */
+export declare function resumeUnkillableBatch(store: SchedStore, spawnDeps: Pick<SpawnDeps, 'isAlive'>, batchId: string, now?: Date): {
+    batch: string;
+    outcome: 'resumed';
+    from: 'agent-unkillable';
+};
 /**
  * The current member's own branch: `batch/<id>-m<n>-<issue>` (#677) — member
  * index and issue both in the name, and deterministic from persisted state,

package/dist/batch-dispatch.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"batch-dispatch.d.ts","sourceRoot":"","sources":["../src/batch-dispatch.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8DG;AAWH,OAAO,EAIL,KAAK,YAAY,EAGjB,KAAK,gBAAgB,EAGtB,MAAM,gBAAgB,CAAC;AAUxB,OAAO,EAeL,KAAK,gBAAgB,EAGrB,KAAK,SAAS,EAEf,MAAM,YAAY,CAAC;AAGpB,OAAO,EAGL,KAAK,WAAW,EAcjB,MAAM,eAAe,CAAC;AACvB,OAAO,EAAE,KAAK,OAAO,EAAa,MAAM,WAAW,CAAC;AAMpD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AAW5C,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,WAAW,CAAC;AAExC,OAAO,EACL,KAAK,iBAAiB,EActB,KAAK,WAAW,EACjB,MAAM,YAAY,CAAC;AAGpB,OAAO,EA2BL,kBAAkB,EAKnB,MAAM,SAAS,CAAC;AACjB,OAAO,EACL,KAAK,QAAQ,EAMd,MAAM,YAAY,CAAC;AACpB,OAAO,KAAK,EAGV,UAAU,EAEV,oBAAoB,EAKpB,kBAAkB,EAClB,cAAc,EAId,WAAW,EACX,UAAU,EAEX,MAAM,SAAS,CAAC;AAajB,YAAY,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AAI1C,OAAO,EAAE,kBAAkB,EAAE,CAAC;AAE9B,8DAA8D;AAC9D,MAAM,WAAW,iBAAiB;IAChC,KAAK,EAAE,UAAU,CAAC;IAClB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,WAAW,CAAC;IACzB,SAAS,EAAE,SAAS,CAAC;IACrB,GAAG,EAAE,MAAM,IAAI,CAAC;IAChB,uFAAuF;IACvF,OAAO,EAAE,MAAM,CAAC;IAChB,0FAA0F;IAC1F,IAAI,EAAE,MAAM,CAAC;IACb;;;;;;OAMG;IACH,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB;;;;;OAKG;IACH,QAAQ,EAAE,CAAC,QAAQ,EAAE,MAAM,EAAE,GAAG,CAAC,EAAE,iBAAiB,KAAK,WAAW,CAAC;IACrE;;;;;;;;;;;;;;;;;;;OAmBG;IACH,aAAa,CAAC,EAAE,CAAC,QAAQ,EAAE,MAAM,EAAE,YAAY,EAAE,MAAM,KAAK,oBAAoB,CAAC;IACjF,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB;;;OAGG;IACH,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB;;;;;;OAMG;IACH,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,8FAA8F;AAC9F,MAAM,WAAW,eAAe;IAC9B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,SAAS,EAAE,MAAM,EAAE,CAAC;IACpB,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,aAAa,EAAE,MAAM,EAAE,CAAC;IACxB,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,uFAAuF;IACvF,qBAAqB,EAAE,MAAM,EAAE,CAAC;IAChC,8FAA8F;IAC9F,OAAO,EAAE,MAAM,EAAE,CAAC;CACnB;AAED;;;;;;GAMG;AACH,MAAM,WAAW,aAAa;IAC5B,MAAM,EAAE,MAAM,CAAC;IACf,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IACjC;;;;OAIG;IACH,IAAI,CAAC,EAAE,cAAc,CAAC;CACvB;AAgtBD;;;;;;;GAOG;AACH,wBAAgB,eAAe,CAAC,OAAO,EAAE,MAAM,EAAE,WAAW,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,CAE3F;AAuuDD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,wBAAgB,sBAAsB,CACpC,IAAI,EAAE,iBAAiB,EACvB,MAAM,EAAE,WAAW,EACnB,QAAQ,EAAE,gBAAgB,EAC1B,OAAO,EAAE,MAAM,EACf,KAAK,EAAE,UAAU,EACjB,WAAW,EAAE,MAAM,EACnB,OAAO,EAAE,aAAa,EACtB,GAAG,EAAE,IAAI,EACT,MAAM,EAAE,eAAe;AACvB,kGAAkG;AAClG,QAAQ,CAAC,EAAE,CAAC,CAAC,EAAE,UAAU,KAAK,UAAU,EACxC,UAAU,CAAC,EAAE,MAAM,IAAI,GACtB,IAAI,CAiBN;AAoWD;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,wBAAgB,iBAAiB,CAC/B,IAAI,EAAE,iBAAiB,EACvB,MAAM,EAAE,WAAW,EACnB,QAAQ,EAAE,gBAAgB,EAC1B,OAAO,EAAE,MAAM,EACf,GAAG,EAAE,IAAI,GACR;IACD,OAAO,EAAE,eAAe,GAAG,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC;IAC/D,UAAU,EAAE,MAAM,CAAC;IACnB;;;;;OAKG;IACH,SAAS,CAAC,EAAE,cAAc,GAAG,qBAAqB,CAAC;IACnD,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB,CA+KA;AAugCD;;;;;;;;GAQG;AACH,eAAO,MAAM,wBAAwB,IAAI,CAAC;AAgoB1C;;;;;;GAMG;AACH,wBAAgB,4BAA4B,CAAC,KAAK,EAAE,UAAU,EAAE,KAAK,EAAE,UAAU,GAAG,MAAM,EAAE,CAuB3F;AA6HD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAwCG;AACH,wBAAgB,4BAA4B,CAC1C,IAAI,EAAE,iBAAiB,EACvB,GAAG,EAAE,IAAI,EACT,MAAM,EAAE,eAAe,EACvB,SAAS,EAAE,iBAAiB,GAAG,SAAS,GACvC,IAAI,CAmMN;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAyCG;AACH,wBAAgB,sBAAsB,CAAC,IAAI,EAAE,iBAAiB,EAAE,GAAG,EAAE,IAAI,GAAG,IAAI,CAuG/E;AAED;;;;;;GAMG;AACH,UAAU,iBAAiB;IACzB,IAAI,EAAE,MAAM,CAAC;IACb,gDAAgD;IAChD,IAAI,EAAE,gBAAgB,CAAC;IACvB,uEAAuE;IACvE,SAAS,EAAE,gBAAgB,CAAC;IAC5B,YAAY,EAAE,YAAY,CAAC;CAC5B;AAoTD;;;;;;;;;;;GAWG;AACH,oCAAoC;AACpC,wBAAgB,qBAAqB,CACnC,KAAK,EAAE,UAAU,EACjB,KAAK,EAAE,UAAU,EACjB,MAAM,EAAE,WAAW,GAClB,kBAAkB,CAQpB;AAgnBD;;;;;;;GAOG;AACH,eAAO,MAAM,qBAAqB,IAAI,CAAC;AA8UvC;;;;;;;;;;;;;GAaG;AACH,wBAAgB,YAAY,CAC1B,IAAI,EAAE,iBAAiB,EACvB,MAAM,EAAE,WAAW,EACnB,QAAQ,EAAE,gBAAgB,GACzB,eAAe,CAySjB"}
+{"version":3,"file":"batch-dispatch.d.ts","sourceRoot":"","sources":["../src/batch-dispatch.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8DG;AAWH,OAAO,EAIL,KAAK,YAAY,EAGjB,KAAK,gBAAgB,EAGtB,MAAM,gBAAgB,CAAC;AAUxB,OAAO,EAeL,KAAK,gBAAgB,EAGrB,KAAK,SAAS,EAEf,MAAM,YAAY,CAAC;AAGpB,OAAO,EAGL,KAAK,WAAW,EAcjB,MAAM,eAAe,CAAC;AACvB,OAAO,EAAE,KAAK,OAAO,EAAa,MAAM,WAAW,CAAC;AAMpD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AAW5C,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,WAAW,CAAC;AAExC,OAAO,EACL,KAAK,iBAAiB,EActB,KAAK,WAAW,EACjB,MAAM,YAAY,CAAC;AAGpB,OAAO,EA2BL,kBAAkB,EAKnB,MAAM,SAAS,CAAC;AACjB,OAAO,EACL,KAAK,QAAQ,EAMd,MAAM,YAAY,CAAC;AACpB,OAAO,KAAK,EAGV,UAAU,EAEV,oBAAoB,EAKpB,kBAAkB,EAClB,cAAc,EAId,WAAW,EACX,UAAU,EAEX,MAAM,SAAS,CAAC;AAajB,YAAY,EAAE,UAAU,EAAE,MAAM,SAAS,CAAC;AAI1C,OAAO,EAAE,kBAAkB,EAAE,CAAC;AAE9B,8DAA8D;AAC9D,MAAM,WAAW,iBAAiB;IAChC,KAAK,EAAE,UAAU,CAAC;IAClB,OAAO,EAAE,OAAO,CAAC;IACjB,WAAW,EAAE,WAAW,CAAC;IACzB,SAAS,EAAE,SAAS,CAAC;IACrB,GAAG,EAAE,MAAM,IAAI,CAAC;IAChB,uFAAuF;IACvF,OAAO,EAAE,MAAM,CAAC;IAChB,0FAA0F;IAC1F,IAAI,EAAE,MAAM,CAAC;IACb;;;;;;OAMG;IACH,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB;;;;;OAKG;IACH,QAAQ,EAAE,CAAC,QAAQ,EAAE,MAAM,EAAE,GAAG,CAAC,EAAE,iBAAiB,KAAK,WAAW,CAAC;IACrE;;;;;;;;;;;;;;;;;;;OAmBG;IACH,aAAa,CAAC,EAAE,CAAC,QAAQ,EAAE,MAAM,EAAE,YAAY,EAAE,MAAM,KAAK,oBAAoB,CAAC;IACjF,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB;;;OAGG;IACH,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB;;;;;;OAMG;IACH,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,8FAA8F;AAC9F,MAAM,WAAW,eAAe;IAC9B,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,SAAS,EAAE,MAAM,EAAE,CAAC;IACpB,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,aAAa,EAAE,MAAM,EAAE,CAAC;IACxB,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,uFAAuF;IACvF,qBAAqB,EAAE,MAAM,EAAE,CAAC;IAChC,8FAA8F;IAC9F,OAAO,EAAE,MAAM,EAAE,CAAC;CACnB;AAED;;;;;;GAMG;AACH,MAAM,WAAW,aAAa;IAC5B,MAAM,EAAE,MAAM,CAAC;IACf,MAAM,EAAE,MAAM,CAAC;IACf,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IACjC;;;;OAIG;IACH,IAAI,CAAC,EAAE,cAAc,CAAC;CACvB;AAyKD;;;;GAIG;AACH,wBAAgB,qBAAqB,CACnC,KAAK,EAAE,UAAU,EACjB,SAAS,EAAE,IAAI,CAAC,SAAS,EAAE,SAAS,CAAC,EACrC,OAAO,EAAE,MAAM,EACf,GAAG,GAAE,IAAiB,GACrB;IAAE,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,SAAS,CAAC;IAAC,IAAI,EAAE,kBAAkB,CAAA;CAAE,CAmBjE;AAyiBD;;;;;;;GAOG;AACH,wBAAgB,eAAe,CAAC,OAAO,EAAE,MAAM,EAAE,WAAW,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,CAE3F;AAuuDD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,wBAAgB,sBAAsB,CACpC,IAAI,EAAE,iBAAiB,EACvB,MAAM,EAAE,WAAW,EACnB,QAAQ,EAAE,gBAAgB,EAC1B,OAAO,EAAE,MAAM,EACf,KAAK,EAAE,UAAU,EACjB,WAAW,EAAE,MAAM,EACnB,OAAO,EAAE,aAAa,EACtB,GAAG,EAAE,IAAI,EACT,MAAM,EAAE,eAAe;AACvB,kGAAkG;AAClG,QAAQ,CAAC,EAAE,CAAC,CAAC,EAAE,UAAU,KAAK,UAAU,EACxC,UAAU,CAAC,EAAE,MAAM,IAAI,GACtB,IAAI,CAiBN;AAoWD;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,wBAAgB,iBAAiB,CAC/B,IAAI,EAAE,iBAAiB,EACvB,MAAM,EAAE,WAAW,EACnB,QAAQ,EAAE,gBAAgB,EAC1B,OAAO,EAAE,MAAM,EACf,GAAG,EAAE,IAAI,GACR;IACD,OAAO,EAAE,eAAe,GAAG,SAAS,GAAG,WAAW,GAAG,SAAS,CAAC;IAC/D,UAAU,EAAE,MAAM,CAAC;IACnB;;;;;OAKG;IACH,SAAS,CAAC,EAAE,cAAc,GAAG,qBAAqB,CAAC;IACnD,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB,CA+KA;AAiiCD;;;;;;;;GAQG;AACH,eAAO,MAAM,wBAAwB,IAAI,CAAC;AAgoB1C;;;;;;GAMG;AACH,wBAAgB,4BAA4B,CAAC,KAAK,EAAE,UAAU,EAAE,KAAK,EAAE,UAAU,GAAG,MAAM,EAAE,CAuB3F;AA6HD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAwCG;AACH,wBAAgB,4BAA4B,CAC1C,IAAI,EAAE,iBAAiB,EACvB,GAAG,EAAE,IAAI,EACT,MAAM,EAAE,eAAe,EACvB,SAAS,EAAE,iBAAiB,GAAG,SAAS,GACvC,IAAI,CAmMN;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAyCG;AACH,wBAAgB,sBAAsB,CAAC,IAAI,EAAE,iBAAiB,EAAE,GAAG,EAAE,IAAI,GAAG,IAAI,CAuG/E;AAED;;;;;;GAMG;AACH,UAAU,iBAAiB;IACzB,IAAI,EAAE,MAAM,CAAC;IACb,gDAAgD;IAChD,IAAI,EAAE,gBAAgB,CAAC;IACvB,uEAAuE;IACvE,SAAS,EAAE,gBAAgB,CAAC;IAC5B,YAAY,EAAE,YAAY,CAAC;CAC5B;AAoTD;;;;;;;;;;;GAWG;AACH,oCAAoC;AACpC,wBAAgB,qBAAqB,CACnC,KAAK,EAAE,UAAU,EACjB,KAAK,EAAE,UAAU,EACjB,MAAM,EAAE,WAAW,GAClB,kBAAkB,CAQpB;AA2nBD;;;;;;;GAOG;AACH,eAAO,MAAM,qBAAqB,IAAI,CAAC;AA8UvC;;;;;;;;;;;;;GAaG;AACH,wBAAgB,YAAY,CAC1B,IAAI,EAAE,iBAAiB,EACvB,MAAM,EAAE,WAAW,EACnB,QAAQ,EAAE,gBAAgB,GACzB,eAAe,CAySjB"}

package/dist/batch-dispatch.js

@@ -97,6 +97,7 @@ var __importStar = (this && this.__importStar) || (function () {
 })();
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.MAX_LAND_PUSH_RETRIES = exports.MAX_BATCH_AGENT_RESPAWNS = exports.tailMembersRefusal = void 0;
+exports.resumeUnkillableBatch = resumeUnkillableBatch;
 exports.memberBranchFor = memberBranchFor;
 exports.evictMemberAndContinue = evictMemberAndContinue;
 exports.resumeBlockedGate = resumeBlockedGate;
@@ -264,6 +265,27 @@ function headTreeOf(deps, worktree) {
 function agentAlive(deps, slot) {
     return slot.pid !== null && deps.spawnDeps.isAlive(slot.pid, slot.pid_start ?? undefined);
 }
+/** #1244: this block originates only in executing's member stop ladder.
+ * Keep slots and member runs so normal reconciliation finishes that exact
+ * decision, rather than redispatching members or discarding landed work.
+ * Liveness and transition share the state lock with dispatch/stop.
+ */
+function resumeUnkillableBatch(store, spawnDeps, batchId, now = new Date()) {
+    return store.withLock((state) => {
+        const batch = (0, state_1.findBatch)(state, batchId);
+        if (batch?.status !== 'blocked' || batch.blocked_reason !== 'agent-unkillable') {
+            throw new types_1.SchedNotFoundError(`Batch ${batchId} is not blocked on agent-unkillable`);
+        }
+        const live = (0, state_1.slotsForBatch)(state, batchId).filter((slot) => slot.pid !== null && spawnDeps.isAlive(slot.pid, slot.pid_start ?? undefined));
+        if (live.length > 0) {
+            throw new types_1.SchedNotFoundError(`Batch ${batchId} still has live agent(s) in slot(s) ${live.map((s) => s.id).join(',')} — wait for them to exit before resuming`);
+        }
+        return {
+            state: (0, state_1.transitionBatch)(state, batchId, 'executing', { blocked_reason: null }, now),
+            result: { batch: batchId, outcome: 'resumed', from: 'agent-unkillable' },
+        };
+    });
+}
 /**
  * Stop the slot's agent if it is still alive (pid-start-safe: a reused pid is
  * never signalled). SIGTERM only — no wait, no escalation. A decision that
@@ -2708,8 +2730,8 @@ function currentDispatchStamp(stamp, spawnedAt, now) {
 }
 /**
  * Stop the slot's agent before a decision that must not run alongside it
- * (#844). Returns `true` while it is still alive — the caller waits and
- * decides on a later tick — and `false` once it is gone (or never had a pid).
+ * (#844). Returns waiting while alive, gone once dead (or without a pid),
+ * and unkillable once SIGKILL's bound expires and the batch blocks.
  *
  * The first sight sends SIGTERM, stamps `kill_sent_at` and journals
  * `member-stop-requested` once; each later tick
@@ -2718,7 +2740,7 @@ function currentDispatchStamp(stamp, spawnedAt, now) {
  * `kill-escalated` once (`kill_escalated_at`). Pre-#844 this path re-sent
  * SIGTERM forever, so an agent ignoring it held its member indefinitely.
  */
-function stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now, result) {
+function stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now, result, reason) {
     if (!agentAlive(deps, slot))
         return 'gone';
     const pid = slot.pid;
@@ -2745,8 +2767,10 @@ function stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now,
                 slot: slot.id,
                 detail: `agent remained alive ${dispatch_1.KILL_ESCALATION_MS / 1000}s after SIGKILL — blocking for operator`,
             });
-            blockBatchForOperator(deps, config, batchId, { reason: 'agent-unkillable' }, now, result);
         }
+        // A crash after the stamp but before the block must finish the block on
+        // the next tick; the stamp deduplicates telemetry, not the transition.
+        blockBatchForOperator(deps, config, batchId, { reason: 'agent-unkillable' }, now, result);
         return 'unkillable';
     }
     const firstEscalation = escalate && currentDispatchStamp(slot.kill_escalated_at, slot.spawned_at, now) === null;
@@ -2768,6 +2792,7 @@ function stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now,
     }
     if (sentAt === null) {
         journalEvent(deps, 'member-stop-requested', unit(batchId), {
+            reason,
             issue: memberIssue,
             pid,
             slot: slot.id,
@@ -2803,7 +2828,7 @@ function repromptWrongProcedure(deps, config, batchId, memberIssue, milestoneAt,
     // dying full-cycle run could still push, post, or open a PR. Signal it and
     // decide on a later tick, once it is gone — and so the re-prompt's
     // `reprompted_at` fence postdates everything it could still post.
-    if (stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now, result) !== 'gone')
+    if (stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now, result, 'wrong-procedure') !== 'gone')
         return 'waiting';
     const outcome = deps.store.withLock((s) => {
         const b = (0, state_1.findBatch)(s, batchId);
@@ -2860,7 +2885,7 @@ function wrongProcedureFailure(lastTool, shippedPr = null) {
 function decideWrongProcedure(deps, config, batchId, memberIssue, milestone, slot, release, lastTool, now, result) {
     const shipped = (0, groundtruth_1.wrongProcedureShippedPr)(milestone);
     if (shipped !== null) {
-        if (stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now, result) !== 'gone')
+        if (stopAgentBeforeDeciding(deps, config, batchId, memberIssue, slot, now, result, 'wrong-procedure') !== 'gone')
             return 'waiting';
         // #844: the caller's eviction releases the slot in its own write.
         return wrongProcedureFailure(lastTool(), shipped);
@@ -4759,7 +4784,7 @@ function reconcileParallelMemberSlot(deps, config, dispatch, batchId, run, slot,
         // The member's tree may be rebased or removed immediately after its slot
         // is released. Keep this slot draining until the agent is really gone so
         // the #844 SIGKILL bound remains available to a SIGTERM-ignoring agent.
-        if (stopAgentBeforeDeciding(deps, config, batchId, run.issue, slot, now, result) !== 'gone')
+        if (stopAgentBeforeDeciding(deps, config, batchId, run.issue, slot, now, result, 'member-complete') !== 'gone')
             return;
         journalEvent(deps, 'external-advance', unit(batchId), {
             issue: run.issue,