@ai-dossier/sched 0.7.1 → 0.9.0

package/dist/batch-dispatch.js

@@ -0,0 +1,1114 @@
+"use strict";
+/**
+ * Batch dispatch (#523, RFC-0001 §C.4/D.2/D.3): the missing driver that
+ * executes `batch:<id>` units. #498 landed the batch failure-recovery library
+ * (attribution/bisect/eviction/dissolve, `recovery.ts`) and the batch state
+ * machine (`state.ts`); readiness/placement already treat a `ready` batch as
+ * a runnable unit (`readiness.ts`, `scheduler.ts`). Nothing dispatched one
+ * until now.
+ *
+ * Shape, mirroring `engine.ts`'s per-issue dispatch: claim a slot → spawn an
+ * agent → poll ground truth → verify → transition. Generalized to `BatchEntry`
+ * at batch-phase granularity instead of per-issue-phase granularity:
+ *
+ * ```
+ * ready → executing(member i/N) ⟲ → validating → reviewing → shipping
+ *   → awaiting-merge → merged → deployed → reported → done
+ * failure rails (RFC F.2/F.8/F.9):
+ *   executing → dissolving  (a member self-reports blocked, RFC F.1)
+ *   validating → attributing → (fixing | evicting) → validating
+ *              → dissolving
+ * ```
+ *
+ * NO batch claim — not the first (`ready → executing`) nor any continuation
+ * (a later member, the tail agent, the report agent, the fix agent) — ever
+ * goes through `computeAssignments`/`runnableUnits`. Every one is a bespoke
+ * free-capacity-gated assignment, the same shape `engine.ts`'s
+ * `dispatchReportAgents` already uses (`runnableUnits` only ever offers a
+ * `status === 'ready'` batch, i.e. the moment BEFORE any claim). Between
+ * steps — a suite run, a PR merge wait — the slot is released to `idle` and
+ * holds no capacity (AC5): only a live member/tail/report/fix agent holds a
+ * slot.
+ *
+ * The aggregate suite itself is deterministic engine work, not an LLM step —
+ * it runs with no slot claimed at all, matching AC5's "member or batch-LLM-step"
+ * wording precisely.
+ *
+ * Two distinct failure rails, deliberately different:
+ * - A member's OWN agent reports itself blocked (its own gate never went
+ *   green) — evicted directly, no attribution needed: the offender is already
+ *   known, and either it has no commits yet (blocked before implementing) or
+ *   its commits are exactly what gets reverted.
+ * - The AGGREGATE suite (run by the engine after every member individually
+ *   went green) comes back red — an integration-level conflict no member's own
+ *   gate caught. THIS is what `recovery.ts`'s attribution/fix/evict pipeline
+ *   exists for (RFC F.2).
+ *
+ * Scope decisions recorded here, not silently cut: no `git bisect` stage for
+ * an ambiguous aggregate failure (bisect needs a per-project "run only these
+ * tests" command this module has no generic way to construct) — an
+ * unattributable red aggregate suite dissolves the batch rather than
+ * bisecting, which `attributing → dissolving` already models. No worktree-pool
+ * integration for batch-setup — cold `git worktree add` only, mirroring
+ * `teardown.ts`'s cold path. No per-phase stall/escalation ladder for batch
+ * sub-agents — a dead-without-verification agent is treated as blocked and
+ * evicted/reported rather than redispatched stronger. Both are documented
+ * follow-ups, not gaps discovered later.
+ */
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.runBatchTick = runBatchTick;
+const path = __importStar(require("node:path"));
+const attribution_1 = require("./attribution");
+const dispatch_1 = require("./dispatch");
+const groundtruth_1 = require("./groundtruth");
+const journal_1 = require("./journal");
+const recovery_1 = require("./recovery");
+const scheduler_1 = require("./scheduler");
+const state_1 = require("./state");
+const teardown_1 = require("./teardown");
+function emptyResult() {
+    return { spawned: [], completed: [], parked: [], mergeAccepted: [], failed: [], blocked: [] };
+}
+function unit(batchId) {
+    return `batch:${batchId}`;
+}
+/**
+ * Sanitize one untrusted string before it lands in persisted state, the
+ * journal, or a `sched status` terminal render (CWE-117/150): a milestone
+ * `reason=` value originates from a GitHub issue comment (anyone who can
+ * comment on a member issue can set it) and `parseMilestoneJson` copies it
+ * verbatim with no charset or length bound. Strips control characters
+ * (including the ANSI escape prefix) and bounds the length.
+ */
+function sanitizeUntrustedText(value) {
+    return (value
+        // biome-ignore lint/suspicious/noControlCharactersInRegex: flattening control characters is the point
+        .replace(/[\u0000-\u001F\u007F]/g, ' ')
+        .slice(0, 200));
+}
+/**
+ * Journal one event. Loosely-typed `extra`, matching `engine.ts`'s own local
+ * `journal()` wrapper — `unitEvent`'s stricter `Omit<JournalEvent, ...>` typing
+ * excess-property-checks an inline object literal (e.g. rejecting `pr`, a key
+ * `JournalEvent` doesn't declare), where a pre-typed `Record<string, unknown>`
+ * value passed through a variable does not.
+ */
+function journalEvent(deps, event, unitId, extra = {}) {
+    deps.journal.append((0, journal_1.unitEvent)(event, unitId, extra), deps.now());
+}
+function slotFor(state, batchId) {
+    return state.slots.find((s) => s.unit === unit(batchId));
+}
+/**
+ * Re-apply a state computed OUTSIDE the lock (by `recovery.ts`'s functions,
+ * which necessarily shell out — `git revert`, `ai-dossier runstate post` —
+ * and so cannot themselves run inside `store.withLock`) onto a FRESHLY
+ * loaded state, touching only `batchId`'s own batch record (plus any new
+ * half-batches a dissolve split created) and the named issues' queue
+ * entries. Anything a concurrent process wrote to `fresh` in the meantime —
+ * `sched enqueue`, `sched abandon`, `sched pause` all take the same
+ * cross-process lock — survives, where blindly returning `computed` wholesale
+ * would have silently clobbered it.
+ */
+function applyBatchAndIssues(fresh, computed, batchId, issues) {
+    const updatedBatch = computed.batches.find((b) => b.id === batchId);
+    const batches = fresh.batches.map((b) => (b.id === batchId && updatedBatch ? updatedBatch : b));
+    // A `halved` dissolve creates new batch ids (`<id>-a`/`<id>-b`) that exist
+    // in `computed` but not yet in `fresh`.
+    for (const cb of computed.batches) {
+        if (!batches.some((b) => b.id === cb.id))
+            batches.push(cb);
+    }
+    const issueSet = new Set(issues);
+    const entries = fresh.entries.map((e) => {
+        if (!issueSet.has(e.issue))
+            return e;
+        return computed.entries.find((ce) => ce.issue === e.issue) ?? e;
+    });
+    return { ...fresh, batches, entries };
+}
+/**
+ * Run the aggregate suite, treating a THROWING runner as a red suite with no
+ * failing tests — `recovery.ts`'s own internal `runSuite` wrapper already
+ * does this for calls that go through `beginAttribution`/`evictMembers`/etc,
+ * but `runValidate`/`reconcileFixSlot` call `deps.runSuite` directly (they
+ * need the result before deciding whether to call into `recovery.ts` at
+ * all), so an unguarded throw there would propagate out of `runBatchTick`
+ * into `tick()`'s own catch — a bare `tick-failed` with no unit id, repeating
+ * every reconcile interval forever since nothing about the batch changed.
+ */
+function safeSuite(deps, batchId, worktree) {
+    try {
+        return deps.runSuite(worktree);
+    }
+    catch (err) {
+        const detail = `suite runner threw: ${err.message}`;
+        journalEvent(deps, 'suite-failed', unit(batchId), { detail });
+        return { ok: false, failing: [], detail };
+    }
+}
+function recoveryDeps(deps, batch, now) {
+    return {
+        exec: deps.exec,
+        repoDir: batch.worktree ?? deps.repoDir,
+        journal: deps.journal,
+        postMilestone: (0, recovery_1.createExecMilestonePoster)(deps.exec, { repoDir: deps.repoDir }),
+        runSuite: batch.worktree !== null ? () => deps.runSuite(batch.worktree) : undefined,
+        now: () => now,
+    };
+}
+/**
+ * The declared edge one step closer to `idle` from each `SlotStatus`
+ * (`state.ts`'s `SLOT_BASE_TRANSITIONS`) — `recovering` has no direct edge to
+ * `idle`, only `running`/`failed`, so it routes through `failed` first;
+ * getting this wrong throws `IllegalTransitionError` inside a lock.
+ */
+const NEXT_TOWARD_IDLE = {
+    idle: null,
+    assigned: 'idle',
+    running: 'exited',
+    exited: 'verifying',
+    verifying: 'complete',
+    complete: 'idle',
+    recovering: 'failed',
+    failed: 'idle',
+};
+/** Release a batch's slot to idle, whatever status it currently holds (mirrors `walkSlotToIdle`). */
+function releaseSlot(state, batchId, now) {
+    let next = state;
+    let slot = slotFor(next, batchId);
+    // Bounded: the longest real walk (recovering → failed → idle, or
+    // running → exited → verifying → complete → idle) is 4 hops.
+    for (let i = 0; i < 8 && slot && slot.status !== 'idle'; i++) {
+        const to = NEXT_TOWARD_IDLE[slot.status];
+        if (to === null)
+            break;
+        next = (0, state_1.transitionSlot)(next, slot.id, to, {}, now);
+        slot = slotFor(next, batchId);
+    }
+    return next;
+}
+// --- Batch setup (ready → executing, member 1) ---
+/**
+ * `batch/<id>-<YYYYMMDD>` off `base_branch`, a worktree at
+ * `<repoDir>/worktrees/batch-<id>-<YYYYMMDD>` (cold git only — no pool
+ * integration in this version, see the module doc), and a fresh runstate run
+ * id minted against the anchor. All-or-nothing: any failed step reports the
+ * step name and nothing is partially recorded on the batch.
+ */
+function runBatchSetup(deps, batch, now) {
+    if (batch.anchor === null)
+        return { ok: false, reason: 'no-anchor' };
+    const date = now.toISOString().slice(0, 10).replaceAll('-', '');
+    const branch = `batch/${batch.id}-${date}`;
+    const worktree = path.join(deps.repoDir, 'worktrees', `batch-${batch.id}-${date}`);
+    if (!attribution_1.SAFE_REF_RE.test(branch) || !attribution_1.SAFE_REF_RE.test(batch.base_branch)) {
+        return { ok: false, reason: 'invalid-branch-name' };
+    }
+    // Defense in depth alongside enqueue.ts's `BATCH_ID_RE` (CWE-22): the batch
+    // id is enqueue-time-validated against path-hostile characters already, but
+    // this is the actual point where it becomes a filesystem path — the same
+    // containment check teardown applies on the way OUT must hold on the way IN.
+    const root = deps.exec('git', ['rev-parse', '--show-toplevel'], deps.repoDir) ?? deps.repoDir;
+    if (!(0, teardown_1.isSafeWorktree)(path.resolve(root), worktree)) {
+        return { ok: false, reason: 'invalid-worktree-path' };
+    }
+    const runId = deps.exec('ai-dossier', ['runstate', 'mint', '--issue', String(batch.anchor)], deps.repoDir);
+    if (runId === null || runId.trim() === '')
+        return { ok: false, reason: 'runstate-mint-failed' };
+    const mintedRunId = runId.trim();
+    if (deps.exec('git', ['fetch', 'origin', '--', batch.base_branch], deps.repoDir) === null) {
+        return { ok: false, reason: 'fetch-failed', runId: mintedRunId };
+    }
+    if (deps.exec('git', ['branch', branch, `origin/${batch.base_branch}`], deps.repoDir) === null) {
+        return { ok: false, reason: 'branch-create-failed', runId: mintedRunId };
+    }
+    if (deps.exec('git', ['push', '-u', 'origin', '--', branch], deps.repoDir) === null) {
+        return { ok: false, reason: 'branch-push-failed', runId: mintedRunId };
+    }
+    if (deps.exec('git', ['worktree', 'add', '--', worktree, branch], deps.repoDir) === null) {
+        return { ok: false, reason: 'worktree-add-failed', runId: mintedRunId };
+    }
+    return { ok: true, branch, worktree, runId: mintedRunId };
+}
+/** Spawn one batch member's `slot-cycle` agent into the slot batch-setup (or a prior member) just released. */
+/**
+ * Drive a member's `QueueEntry` through the D.1 slot-line states it must pass
+ * through before `shipped-in-batch` becomes a legal edge (`validated` is the
+ * only state that transitions there) — `classified → batched → waiting →
+ * in-work`, each a no-op waypoint from the batch's perspective (the real
+ * waiting/working happens at BATCH granularity), applied idempotently so a
+ * member already past a given waypoint is left alone.
+ */
+function advanceMemberToInWork(state, memberIssue, now) {
+    let next = state;
+    const chain = [
+        ['queued', 'classified'],
+        ['classified', 'batched'],
+        ['batched', 'waiting'],
+        ['waiting', 'in-work'],
+    ];
+    for (const [from, to] of chain) {
+        if ((0, state_1.findEntry)(next, memberIssue)?.status === from) {
+            next = (0, state_1.transitionIssue)(next, memberIssue, to, {}, now);
+        }
+    }
+    return next;
+}
+/** The completion half of the same chain: `in-work → committed → validated` (see `advanceMemberToInWork`). */
+function advanceMemberToValidated(state, memberIssue, now) {
+    let next = state;
+    const chain = [
+        ['in-work', 'committed'],
+        ['committed', 'validated'],
+    ];
+    for (const [from, to] of chain) {
+        if ((0, state_1.findEntry)(next, memberIssue)?.status === from) {
+            next = (0, state_1.transitionIssue)(next, memberIssue, to, {}, now);
+        }
+    }
+    return next;
+}
+function spawnMember(deps, dispatch, state, slot, batchId, now, result) {
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch || batch.worktree === null) {
+        // A leaked `assigned` slot with `pid: null` is invisible to `dead`
+        // detection (nothing ever kills/reclaims it) — release it here rather
+        // than leaving the batch permanently down one slot of capacity.
+        journalEvent(deps, 'unit-failed', unit(batchId), {
+            reason: 'no-worktree',
+            detail: 'spawnMember: batch has no worktree — batch-setup has not landed',
+        });
+        return releaseSlot(state, batchId, now);
+    }
+    const memberIssue = batch.members[batch.executing_member - 1];
+    if (memberIssue === undefined) {
+        journalEvent(deps, 'unit-failed', unit(batchId), {
+            reason: 'no-member',
+            detail: `spawnMember: executing_member ${batch.executing_member} has no member issue`,
+        });
+        return releaseSlot(state, batchId, now);
+    }
+    const withStatus = advanceMemberToInWork(state, memberIssue, now);
+    const tier = (0, state_1.findEntry)(withStatus, memberIssue)?.tier ?? 'mid';
+    const cmd = (0, dispatch_1.buildAgentCommand)(dispatch.command, tier, memberIssue, dispatch.tierModels);
+    const prompt = (0, dispatch_1.buildMemberPrompt)(dispatch.memberPrompt, memberIssue, batchId, batch.worktree);
+    const logFile = path.join(deps.store.runsDir, `${(0, dispatch_1.unitLogName)(unit(batchId))}-m${batch.executing_member}-${memberIssue}.log`);
+    let pid;
+    try {
+        pid = deps.spawnDeps.spawn(cmd, prompt, logFile);
+    }
+    catch (err) {
+        journalEvent(deps, 'unit-failed', unit(batchId), {
+            issue: memberIssue,
+            reason: 'spawn-error',
+            detail: `member #${memberIssue} spawn failed: ${err.message}`,
+        });
+        result.failed.push(unit(batchId));
+        return releaseSlot(withStatus, batchId, now);
+    }
+    const patch = {
+        pid,
+        pid_start: deps.spawnDeps.processStart(pid),
+        phase: 'member',
+        last_progress_at: now.toISOString(),
+    };
+    const next = slot.status === 'assigned' || slot.status === 'recovering'
+        ? (0, state_1.transitionSlot)(withStatus, slot.id, 'running', patch, now)
+        : withStatus;
+    deps.journal.append((0, journal_1.unitEvent)('spawned', unit(batchId), {
+        pid,
+        tier,
+        slot: slot.id,
+        issue: memberIssue,
+        detail: `member ${batch.executing_member}/${batch.members.length}`,
+    }), now);
+    result.spawned.push(unit(batchId));
+    return next;
+}
+/**
+ * First claim of a `ready` batch: assign it a slot, run batch-setup (real
+ * network round trips — `ai-dossier runstate mint`, `git fetch`/`push`), land
+ * the results, then spawn member 1 in the SAME slot rather than releasing and
+ * re-claiming: setup already holds the slot for its own duration, and
+ * splitting it into two capacity-gated claims would only add a second gate
+ * for no benefit — the slot is going to member 1 immediately either way.
+ */
+function claimAndSetup(deps, config, dispatch, batchId, now, result) {
+    const claimedSlot = deps.store.withLock((state) => {
+        const batch = (0, state_1.findBatch)(state, batchId);
+        if (!batch || batch.status !== 'ready' || batch.anchor === null || slotFor(state, batchId)) {
+            return { state, result: null };
+        }
+        if ((0, scheduler_1.freeCapacity)(state, config) === 0)
+            return { state, result: null };
+        const assigned = (0, scheduler_1.assignToIdleSlot)(state, unit(batchId), 'batch-setup', now);
+        return { state: assigned.state, result: assigned.slotId };
+    });
+    if (claimedSlot === null)
+        return;
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch)
+        return;
+    const setup = runBatchSetup(deps, batch, now);
+    const poster = (0, recovery_1.createExecMilestonePoster)(deps.exec, { repoDir: deps.repoDir });
+    if (!setup.ok) {
+        // `ai-dossier runstate post` REQUIRES a run id (types.ts's `BatchEntry.run_id`
+        // doc) — posting with an empty string silently fails the CLI call. Only post
+        // when the mint step actually landed one (`setup.runId`, when a LATER step
+        // failed) or the batch already carries one from an earlier attempt.
+        const runId = setup.runId ?? batch.run_id;
+        if (batch.anchor !== null && runId !== null) {
+            poster(batch.anchor, runId, {
+                phase: 'batch-setup',
+                status: 'blocked',
+                kv: { reason: setup.reason },
+            });
+        }
+        else {
+            journalEvent(deps, 'milestone-post-failed', unit(batchId), {
+                detail: `batch-setup blocked (${setup.reason}) — no run id to post to yet`,
+            });
+        }
+        deps.journal.append((0, journal_1.unitEvent)('batch-setup-failed', unit(batchId), { detail: setup.reason }), now);
+        result.failed.push(unit(batchId));
+        deps.store.withLock((s) => ({ state: releaseSlot(s, batchId, now), result: undefined }));
+        return;
+    }
+    poster(batch.anchor, setup.runId, {
+        phase: 'batch-setup',
+        status: 'done',
+        kv: { branch: setup.branch, worktree: setup.worktree, base_branch: batch.base_branch },
+    });
+    deps.journal.append((0, journal_1.unitEvent)('batch-setup-done', unit(batchId), { detail: setup.worktree }), now);
+    deps.store.withLock((s) => {
+        const b = (0, state_1.findBatch)(s, batchId);
+        if (!b || b.status !== 'ready') {
+            // The batch moved (dissolved/abandoned) between the claim and here — a
+            // real worktree now exists that nothing else knows about, and the slot
+            // this claim took is still `assigned` with no agent in it. Release the
+            // slot rather than leaking capacity; the worktree is orphaned (named in
+            // the journal for manual cleanup — it is not this rare-race path's job
+            // to guess whether reusing or removing it is safe).
+            journalEvent(deps, 'unit-failed', unit(batchId), {
+                reason: 'batch-left-ready-during-setup',
+                detail: `worktree ${setup.worktree} created but batch is now '${b?.status ?? 'gone'}' — orphaned, manual cleanup required`,
+            });
+            return { state: releaseSlot(s, batchId, now), result: undefined };
+        }
+        let next = (0, state_1.patchBatch)(s, batchId, { branch: setup.branch, worktree: setup.worktree, run_id: setup.runId }, now);
+        next = (0, state_1.transitionBatch)(next, batchId, 'executing', { executing_member: 1 }, now);
+        const slot = slotFor(next, batchId);
+        if (slot)
+            next = spawnMember(deps, dispatch, next, slot, batchId, now, result);
+        return { state: next, result: undefined };
+    });
+}
+// --- Continuation: claim a fresh slot for the next live step ---
+/**
+ * Claim a fresh idle slot for the batch's next live step (a later member, the
+ * tail agent, the report agent), gated on free capacity exactly like
+ * `dispatchReportAgents` — never through `computeAssignments`/`runnableUnits`
+ * again (those only ever offer a `status === 'ready'` batch).
+ */
+function claimAndSpawn(deps, config, batchId, phase, now, spawn) {
+    return deps.store.withLock((state) => {
+        const batch = (0, state_1.findBatch)(state, batchId);
+        if (!batch || slotFor(state, batchId))
+            return { state, result: false };
+        if ((0, scheduler_1.freeCapacity)(state, config) === 0)
+            return { state, result: false };
+        const assigned = (0, scheduler_1.assignToIdleSlot)(state, unit(batchId), phase, now, 'cycle');
+        const slot = assigned.state.slots.find((s) => s.id === assigned.slotId);
+        if (!slot)
+            return { state, result: false };
+        return { state: spawn(assigned.state, slot), result: true };
+    });
+}
+function spawnMemberContinuation(deps, config, dispatch, batchId, now, result) {
+    claimAndSpawn(deps, config, batchId, 'member', now, (state, slot) => spawnMember(deps, dispatch, state, slot, batchId, now, result));
+}
+function spawnTailAgent(deps, config, dispatch, batchId, now, result) {
+    claimAndSpawn(deps, config, batchId, 'reviewing', now, (state, slot) => {
+        const batch = (0, state_1.findBatch)(state, batchId);
+        if (!batch || batch.worktree === null || batch.anchor === null) {
+            journalEvent(deps, 'unit-failed', unit(batchId), {
+                reason: 'no-worktree-or-anchor',
+                detail: 'spawnTailAgent: batch has no worktree/anchor',
+            });
+            return releaseSlot(state, batchId, now);
+        }
+        const cmd = (0, dispatch_1.buildAgentCommand)(dispatch.command, 'strong', batch.anchor, dispatch.tierModels);
+        const prompt = (0, dispatch_1.buildBatchTailPrompt)(dispatch.batchTailPrompt, batchId, batch.anchor, batch.members, batch.worktree);
+        const logFile = path.join(deps.store.runsDir, `${(0, dispatch_1.unitLogName)(unit(batchId))}-tail.log`);
+        let pid;
+        try {
+            pid = deps.spawnDeps.spawn(cmd, prompt, logFile);
+        }
+        catch (err) {
+            journalEvent(deps, 'unit-failed', unit(batchId), {
+                reason: 'spawn-error',
+                detail: `tail agent spawn failed: ${err.message}`,
+            });
+            result.failed.push(unit(batchId));
+            return releaseSlot(state, batchId, now);
+        }
+        const patch = {
+            pid,
+            pid_start: deps.spawnDeps.processStart(pid),
+            phase: 'reviewing',
+            last_progress_at: now.toISOString(),
+        };
+        const next = slot.status === 'assigned' ? (0, state_1.transitionSlot)(state, slot.id, 'running', patch, now) : state;
+        deps.journal.append((0, journal_1.unitEvent)('spawned', unit(batchId), { pid, slot: slot.id }), now);
+        result.spawned.push(unit(batchId));
+        return next;
+    });
+}
+function spawnReportAgent(deps, config, dispatch, batchId, now, result) {
+    claimAndSpawn(deps, config, batchId, 'report', now, (state, slot) => {
+        const batch = (0, state_1.findBatch)(state, batchId);
+        if (!batch || batch.anchor === null) {
+            journalEvent(deps, 'report-failed', unit(batchId), {
+                detail: 'spawnReportAgent: batch has no anchor',
+            });
+            return releaseSlot(state, batchId, now);
+        }
+        const prNumber = batch.pr;
+        if (prNumber === null) {
+            journalEvent(deps, 'report-failed', unit(batchId), {
+                detail: 'spawnReportAgent: batch has no parked pr recorded',
+            });
+            return releaseSlot(state, batchId, now);
+        }
+        const cmd = (0, dispatch_1.buildAgentCommand)(dispatch.command, 'mechanical', batch.anchor, dispatch.tierModels);
+        const prompt = (0, dispatch_1.buildBatchReportPrompt)(dispatch.batchReportPrompt, batchId, batch.anchor, prNumber);
+        const logFile = path.join(deps.store.runsDir, `${(0, dispatch_1.unitLogName)(unit(batchId))}-report.log`);
+        let pid;
+        try {
+            pid = deps.spawnDeps.spawn(cmd, prompt, logFile);
+        }
+        catch (err) {
+            deps.journal.append((0, journal_1.unitEvent)('report-failed', unit(batchId), { detail: err.message }), now);
+            return releaseSlot(state, batchId, now);
+        }
+        const patchState = {
+            pid,
+            pid_start: deps.spawnDeps.processStart(pid),
+            phase: 'report',
+            last_progress_at: now.toISOString(),
+        };
+        const next = slot.status === 'assigned'
+            ? (0, state_1.transitionSlot)(state, slot.id, 'running', patchState, now)
+            : state;
+        journalEvent(deps, 'report-dispatched', unit(batchId), { pid, slot: slot.id, pr: prNumber });
+        result.spawned.push(unit(batchId));
+        return next;
+    });
+}
+// --- Aggregate validate + attribution/fix/evict (RFC F.2) ---
+function memberFootprints(deps, batch) {
+    if (batch.worktree === null)
+        return [];
+    return batch.ranges.map((range) => {
+        // `range.commits` is persisted state — validate as shas before they
+        // become git argv (CWE-88), the same discipline `recovery.ts`'s revert
+        // path applies to the identical values (attribution.ts's `SHA_RE` doc).
+        const commits = range.commits.filter((c) => attribution_1.SHA_RE.test(c));
+        if (commits.length === 0)
+            return { issue: range.issue, changedPaths: [], focusedTests: [] };
+        const out = deps.exec('git', ['show', '--name-only', '--format=', ...commits], batch.worktree);
+        const changedPaths = (out ?? '')
+            .split('\n')
+            .map((l) => l.trim())
+            .filter((l) => l.length > 0);
+        return { issue: range.issue, changedPaths, focusedTests: [] };
+    });
+}
+function boundaryCommits(deps, batch) {
+    if (batch.worktree === null || batch.branch === null)
+        return [];
+    const out = deps.exec('git', ['log', '--reverse', '--format=%H%x09%s', `origin/${batch.base_branch}..${batch.branch}`], batch.worktree);
+    return (0, attribution_1.parseBoundaryCommits)(out);
+}
+/**
+ * `validating`, no live slot: run the aggregate suite (deterministic — no
+ * agent, no slot claimed, matching AC5's "member or batch-LLM-step" wording).
+ * Green proceeds to the tail; red attributes and either fixes one offender or
+ * dissolves when nothing could be attributed (RFC F.2/F.8).
+ */
+function runValidate(deps, config, dispatch, batchId, now, result) {
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch || batch.worktree === null)
+        return;
+    const suite = safeSuite(deps, batchId, batch.worktree);
+    const rDeps = recoveryDeps(deps, batch, now);
+    const poster = (0, recovery_1.createExecMilestonePoster)(deps.exec, { repoDir: deps.repoDir });
+    if (suite.ok) {
+        if (batch.anchor !== null && batch.run_id !== null) {
+            poster(batch.anchor, batch.run_id, { phase: 'batch-validate', status: 'done', kv: {} });
+        }
+        deps.journal.append((0, journal_1.unitEvent)('verify-complete', unit(batchId), { detail: 'suite green' }), now);
+        deps.store.withLock((s) => {
+            const b = (0, state_1.findBatch)(s, batchId);
+            if (!b || b.status !== 'validating')
+                return { state: s, result: undefined };
+            return { state: (0, state_1.transitionBatch)(s, batchId, 'reviewing', {}, now), result: undefined };
+        });
+        spawnTailAgent(deps, config, dispatch, batchId, now, result);
+        return;
+    }
+    const { state: attributed, outcome } = (0, recovery_1.beginAttribution)(state, batchId, { failing: suite.failing, footprints: memberFootprints(deps, batch) }, rDeps);
+    if (outcome.offenders.length === 0) {
+        const dissolve = (0, recovery_1.dissolveBatch)(attributed, batchId, { strategy: 'full', reason: 'unattributable-suite-failure' }, rDeps);
+        deps.store.withLock((s) => ({
+            state: applyBatchAndIssues(s, dissolve.state, batchId, dissolve.requeued),
+            result: undefined,
+        }));
+        teardownBatch(deps, batchId);
+        result.blocked.push(...dissolve.requeued);
+        result.failed.push(unit(batchId));
+        return;
+    }
+    const offender = outcome.offenders[0];
+    const { state: fixing, dispatch: fixDispatch } = (0, recovery_1.beginFixAttempt)(attributed, batchId, offender, rDeps, { config, tests: outcome.attributed.get(offender) ?? [] });
+    deps.store.withLock((s) => ({
+        state: applyBatchAndIssues(s, fixing, batchId, []),
+        result: undefined,
+    }));
+    if (fixDispatch === null) {
+        // Already had its one attempt — evict directly (mirrors the module's own
+        // documented next step when `beginFixAttempt` refuses).
+        evictOffender(deps, config, batchId, offender, outcome.method, now, result);
+        return;
+    }
+    claimAndSpawn(deps, config, batchId, 'fixing', now, (s, slot) => {
+        const logFile = path.join(deps.store.runsDir, `${(0, dispatch_1.unitLogName)(unit(batchId))}-fix-${offender}.log`);
+        let pid;
+        try {
+            pid = deps.spawnDeps.spawn(fixDispatch.command, fixDispatch.prompt, logFile);
+        }
+        catch (err) {
+            journalEvent(deps, 'unit-failed', unit(batchId), {
+                issue: offender,
+                reason: 'fix-spawn-error',
+                detail: `fix agent spawn failed: ${err.message}`,
+            });
+            // The fix attempt was already recorded `dispatched` by `beginFixAttempt`
+            // — a spawn failure never dispatched anything, so resolve it `red`
+            // (pure, no I/O — safe inside this lock) rather than leaving the state
+            // claiming an attempt is in flight forever.
+            const resolved = (0, recovery_1.resolveFixAttempt)(s, batchId, offender, 'red', rDeps).state;
+            return releaseSlot(resolved, batchId, now);
+        }
+        const patch = {
+            pid,
+            pid_start: deps.spawnDeps.processStart(pid),
+            phase: 'fixing',
+            last_progress_at: now.toISOString(),
+        };
+        const next = slot.status === 'assigned' ? (0, state_1.transitionSlot)(s, slot.id, 'running', patch, now) : s;
+        result.spawned.push(unit(batchId));
+        return next;
+    });
+}
+function evictOffender(deps, _config, batchId, offender, attribution, now, result) {
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch)
+        return;
+    const rDeps = recoveryDeps(deps, batch, now);
+    const outcome = (0, recovery_1.evictMembers)(state, batchId, { issues: [offender], reason: 'suite-red-after-fix', attribution, ranges: batch.ranges }, rDeps);
+    deps.store.withLock((s) => ({
+        state: applyBatchAndIssues(s, outcome.state, batchId, outcome.requeued),
+        result: undefined,
+    }));
+    if (outcome.dissolved) {
+        result.failed.push(unit(batchId));
+        return;
+    }
+    if (outcome.suite?.ok) {
+        deps.store.withLock((s) => {
+            const b = (0, state_1.findBatch)(s, batchId);
+            if (!b || b.status !== 'validating')
+                return { state: s, result: undefined };
+            return { state: (0, state_1.transitionBatch)(s, batchId, 'reviewing', {}, now), result: undefined };
+        });
+    }
+}
+// --- Reconcile a batch currently holding a live/exited slot ---
+/**
+ * After a member leaves `executing` (green, gate-failed, or self-blocked):
+ * advance to the next member, or — on the last member — transition to
+ * `validating` and run the aggregate suite. Shared by every exit from
+ * `reconcileMemberSlot` so the pointer-advance rail exists once.
+ */
+function advanceMemberOrValidate(deps, config, dispatch, batchId, memberCount, currentMember, memberIssue, now, result) {
+    const isLast = currentMember >= memberCount;
+    if (isLast) {
+        deps.store.withLock((s) => {
+            const b = (0, state_1.findBatch)(s, batchId);
+            if (!b || b.status !== 'executing')
+                return { state: s, result: undefined };
+            return { state: (0, state_1.transitionBatch)(s, batchId, 'validating', {}, now), result: undefined };
+        });
+        runValidate(deps, config, dispatch, batchId, now, result);
+        return;
+    }
+    deps.store.withLock((s) => {
+        const b = (0, state_1.findBatch)(s, batchId);
+        if (!b || b.status !== 'executing')
+            return { state: s, result: undefined };
+        return {
+            state: (0, state_1.transitionBatch)(s, batchId, 'executing', { executing_member: b.executing_member + 1 }, now),
+            result: undefined,
+        };
+    });
+    journalEvent(deps, 'member-advanced', unit(batchId), { issue: memberIssue });
+    spawnMemberContinuation(deps, config, dispatch, batchId, now, result);
+}
+/**
+ * Evict the current member and either dissolve, or continue the batch via
+ * `advanceMemberOrValidate` — the shared tail of both member-failure rails
+ * (self-reported blocked, and the incremental gate below).
+ */
+function evictMemberAndContinue(deps, config, dispatch, batchId, batch, memberIssue, reason, now, result) {
+    const dissolved = evictMemberDirectly(deps, batchId, memberIssue, reason, now);
+    if (dissolved) {
+        result.failed.push(unit(batchId));
+        return;
+    }
+    advanceMemberOrValidate(deps, config, dispatch, batchId, batch.members.length, batch.executing_member, memberIssue, now, result);
+}
+function reconcileMemberSlot(deps, config, dispatch, batchId, slot, now, result) {
+    const state0 = deps.store.load();
+    const batch = (0, state_1.findBatch)(state0, batchId);
+    if (!batch)
+        return;
+    const memberIssue = batch.members[batch.executing_member - 1];
+    if (memberIssue === undefined)
+        return;
+    const dead = slot.pid !== null && !deps.spawnDeps.isAlive(slot.pid, slot.pid_start ?? undefined);
+    const milestone = deps.groundTruth.latestMilestone(memberIssue);
+    if (milestone === undefined)
+        return; // unreachable — pause this batch's decisions
+    if ((0, groundtruth_1.isMemberComplete)(milestone)) {
+        deps.journal.append((0, journal_1.unitEvent)('external-advance', unit(batchId), {
+            issue: memberIssue,
+            detail: 'member review done',
+        }), now);
+        // Incremental gate (#523 AC2): typecheck + focused tests via `cap run`,
+        // when the repo has a manifest for them — a second, independent check
+        // that the member's own self-reported "done" is real, matching this
+        // codebase's "never trust a claimed completion" ethos (AC2/#464's
+        // `isVerifiedComplete`). `task-failed` evicts the member directly, same
+        // rail as a self-reported block (RFC F.1) — no aggregate suite has run
+        // yet, so there is nothing to attribute.
+        if (batch.worktree !== null && deps.runCapability) {
+            const gateFailure = ['typecheck.run', 'test.focused']
+                .map((id) => ({ id, outcome: deps.runCapability?.(batch.worktree, id) }))
+                .find((r) => r.outcome === 'task-failed');
+            if (gateFailure) {
+                const reason = `incremental-gate-failed:${gateFailure.id}`;
+                journalEvent(deps, 'unit-failed', unit(batchId), {
+                    issue: memberIssue,
+                    reason,
+                    detail: `cap run ${gateFailure.id} reported task-failed after member review done`,
+                });
+                deps.store.withLock((s) => ({ state: releaseSlot(s, batchId, now), result: undefined }));
+                evictMemberAndContinue(deps, config, dispatch, batchId, batch, memberIssue, reason, now, result);
+                return;
+            }
+            // `ok` / `automation-broken` / `capability-unavailable` all proceed —
+            // only a definite task failure blocks a member here.
+        }
+        // The commit-range recompute (`git log`) is a blocking subprocess call —
+        // it must run OUTSIDE the lock, like every other exec in this module;
+        // the result then lands as a pure data patch under the lock (Convention
+        // review: `recordRanges` used to run `git log` INSIDE the withLock
+        // mutator, which is exactly what `engine.ts`'s own "a slow git call never
+        // holds the lock" invariant exists to prevent).
+        const ranges = (0, attribution_1.memberRanges)(boundaryCommits(deps, batch));
+        deps.store.withLock((s) => {
+            let n = releaseSlot(s, batchId, now);
+            n = (0, state_1.patchBatch)(n, batchId, { ranges }, now);
+            n = advanceMemberToValidated(n, memberIssue, now);
+            return { state: n, result: undefined };
+        });
+        result.completed.push(unit(batchId));
+        advanceMemberOrValidate(deps, config, dispatch, batchId, batch.members.length, batch.executing_member, memberIssue, now, result);
+        return;
+    }
+    if ((0, groundtruth_1.isMemberBlocked)(milestone) || dead) {
+        const rawReason = milestone?.keys.reason;
+        const reason = typeof rawReason === 'string' && rawReason.length > 0
+            ? sanitizeUntrustedText(rawReason)
+            : dead
+                ? 'agent-exited-unverified'
+                : 'member-blocked';
+        journalEvent(deps, 'unit-failed', unit(batchId), {
+            issue: memberIssue,
+            reason,
+            detail: 'member blocked',
+        });
+        deps.store.withLock((s) => ({ state: releaseSlot(s, batchId, now), result: undefined }));
+        // A member that never went green (RFC F.1) evicts DIRECTLY — no aggregate
+        // suite has run yet, so there is nothing for `attributing`/`evicting` (the
+        // AGGREGATE-suite-red pipeline, RFC F.2) to attribute or revert: the
+        // offender is already known, and `batch.ranges` has no entry for a member
+        // that never reached `isMemberComplete`. `executing → validating →
+        // attributing → evicting` is not even a legal edge from mid-`executing`
+        // (BATCH_TRANSITIONS has no `executing → evicting`) — this stays entirely
+        // within `executing`/`dissolving`, both of which ARE legal from here.
+        evictMemberAndContinue(deps, config, dispatch, batchId, batch, memberIssue, reason, now, result);
+    }
+}
+/**
+ * Requeue a member full-cycle, record the eviction, and dissolve if this tips
+ * the batch past the ⅓ threshold (RFC F.1/F.8) — WITHOUT going through
+ * `recovery.ts`'s `evictMembers` (which needs `attributing`/`evicting` status
+ * and a commit range to revert; a member evicted here has neither). Returns
+ * whether the batch dissolved.
+ */
+function evictMemberDirectly(deps, batchId, memberIssue, reason, now) {
+    // Pass 1 (pure — requeue + record the eviction): safe to run entirely
+    // inside the lock, unlike `dissolveBatch` below, which shells out
+    // (`deps.exec`/`postMilestone`) and so must NOT hold the lock while it runs.
+    const triggered = deps.store.withLock((s) => {
+        const b = (0, state_1.findBatch)(s, batchId);
+        if (!b)
+            return { state: s, result: false };
+        const evidence = {
+            batch: batchId,
+            reason,
+            failing_tests: [],
+            attribution: 'none',
+            reverted_commits: [],
+            at: now.toISOString(),
+        };
+        const requeueResult = (0, state_1.requeueMember)(s, memberIssue, { mode: 'full', batch: null }, reason, now, { failure_evidence: evidence });
+        let next = requeueResult.state;
+        next = (0, state_1.patchBatch)(next, batchId, {
+            evictions: [
+                ...b.evictions,
+                {
+                    issue: memberIssue,
+                    reason,
+                    attribution: 'none',
+                    reverted_commits: [],
+                    group: [],
+                    at: now.toISOString(),
+                },
+            ],
+        }, now);
+        const updated = (0, state_1.findBatch)(next, batchId);
+        return { state: next, result: updated !== undefined && (0, recovery_1.checkDissolveTrigger)(updated) };
+    });
+    if (!triggered)
+        return false;
+    // Pass 2 (outside the lock — dissolveBatch shells out): re-load fresh
+    // (pass 1's write already landed), dissolve, then re-apply just this
+    // batch's + the requeued members' state under a fresh lock.
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch)
+        return false;
+    const rDeps = recoveryDeps(deps, batch, now);
+    const outcome = (0, recovery_1.dissolveBatch)(state, batchId, { strategy: 'full', reason: 'eviction-threshold' }, rDeps);
+    deps.store.withLock((s) => ({
+        state: applyBatchAndIssues(s, outcome.state, batchId, outcome.requeued),
+        result: undefined,
+    }));
+    teardownBatch(deps, batchId);
+    return true;
+}
+function reconcileFixSlot(deps, config, batchId, slot, now, result) {
+    if (slot.pid !== null && deps.spawnDeps.isAlive(slot.pid, slot.pid_start ?? undefined))
+        return; // still running
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch || batch.worktree === null)
+        return;
+    const offenderRecord = [...batch.fix_attempts].reverse().find((a) => a.outcome === 'dispatched');
+    if (!offenderRecord)
+        return;
+    deps.store.withLock((s) => ({ state: releaseSlot(s, batchId, now), result: undefined }));
+    const suite = safeSuite(deps, batchId, batch.worktree);
+    const rDeps = recoveryDeps(deps, batch, now);
+    const { state: resolved } = (0, recovery_1.resolveFixAttempt)(deps.store.load(), batchId, offenderRecord.issue, suite.ok ? 'green' : 'red', rDeps);
+    deps.store.withLock((s) => ({
+        state: applyBatchAndIssues(s, resolved, batchId, []),
+        result: undefined,
+    }));
+    if (suite.ok) {
+        deps.store.withLock((s) => {
+            const b = (0, state_1.findBatch)(s, batchId);
+            if (!b || b.status !== 'validating')
+                return { state: s, result: undefined };
+            return { state: (0, state_1.transitionBatch)(s, batchId, 'reviewing', {}, now), result: undefined };
+        });
+        return;
+    }
+    evictOffender(deps, config, batchId, offenderRecord.issue, 'overlap', now, result);
+}
+function reconcileTailSlot(deps, batchId, slot, now, result) {
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch || batch.anchor === null)
+        return;
+    const dead = slot.pid !== null && !deps.spawnDeps.isAlive(slot.pid, slot.pid_start ?? undefined);
+    const milestone = deps.groundTruth.latestMilestone(batch.anchor);
+    if (milestone === undefined)
+        return;
+    if (batch.status === 'reviewing' && (0, groundtruth_1.isBatchPhaseDone)(milestone, 'batch-review')) {
+        deps.store.withLock((s) => {
+            const b = (0, state_1.findBatch)(s, batchId);
+            if (!b || b.status !== 'reviewing')
+                return { state: s, result: undefined };
+            return { state: (0, state_1.transitionBatch)(s, batchId, 'shipping', {}, now), result: undefined };
+        });
+        return;
+    }
+    if ((0, groundtruth_1.isBatchTailParked)(milestone)) {
+        const pr = (0, groundtruth_1.prOfMilestone)(milestone);
+        journalEvent(deps, 'pr-parked', unit(batchId), { pr: pr ?? undefined });
+        deps.store.withLock((s) => {
+            let n = releaseSlot(s, batchId, now);
+            const b = (0, state_1.findBatch)(n, batchId);
+            if (!b)
+                return { state: n, result: undefined };
+            n = b.status === 'reviewing' ? (0, state_1.transitionBatch)(n, batchId, 'shipping', {}, now) : n;
+            n = (0, state_1.transitionBatch)(n, batchId, 'awaiting-merge', { pr }, now);
+            return { state: n, result: undefined };
+        });
+        result.parked.push(unit(batchId));
+        return;
+    }
+    if (dead) {
+        deps.journal.append((0, journal_1.unitEvent)('unit-failed', unit(batchId), { reason: 'tail-agent-exited-unverified' }), now);
+        deps.store.withLock((s) => ({ state: releaseSlot(s, batchId, now), result: undefined }));
+        result.failed.push(unit(batchId));
+    }
+}
+function reconcileReportSlot(deps, batchId, slot, now, result) {
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch || batch.anchor === null)
+        return;
+    const dead = slot.pid !== null && !deps.spawnDeps.isAlive(slot.pid, slot.pid_start ?? undefined);
+    const milestone = deps.groundTruth.latestMilestone(batch.anchor);
+    if (milestone === undefined)
+        return;
+    if ((0, groundtruth_1.isBatchPhaseDone)(milestone, 'batch-report')) {
+        deps.journal.append((0, journal_1.unitEvent)('external-advance', unit(batchId), { detail: 'batch report done' }), now);
+        deps.store.withLock((s) => {
+            let n = releaseSlot(s, batchId, now);
+            const b = (0, state_1.findBatch)(n, batchId);
+            if (!b || b.status !== 'deployed')
+                return { state: n, result: undefined };
+            n = (0, state_1.transitionBatch)(n, batchId, 'reported', {}, now);
+            n = (0, state_1.transitionBatch)(n, batchId, 'done', {}, now);
+            return { state: n, result: undefined };
+        });
+        result.completed.push(unit(batchId));
+        teardownBatch(deps, batchId);
+        return;
+    }
+    if (dead) {
+        deps.journal.append((0, journal_1.unitEvent)('report-failed', unit(batchId), { detail: 'unverified exit' }), now);
+        deps.store.withLock((s) => ({ state: releaseSlot(s, batchId, now), result: undefined }));
+    }
+}
+// --- PR watch for `awaiting-merge` batches (mirrors `pollParkedPrs`/`reconcileParked`) ---
+function reconcilePrWatch(deps, now, result) {
+    const state = deps.store.load();
+    for (const batch of state.batches) {
+        if (batch.status !== 'awaiting-merge')
+            continue;
+        const pr = batch.pr;
+        if (pr === null)
+            continue;
+        const truth = deps.groundTruth.prState(pr);
+        if (truth === undefined)
+            continue; // unreachable — keep watching
+        if (truth.state === 'MERGED' && truth.mergedAt !== null) {
+            journalEvent(deps, 'merge-accepted', unit(batch.id), { pr });
+            deps.store.withLock((s) => {
+                const b = (0, state_1.findBatch)(s, batch.id);
+                if (!b || b.status !== 'awaiting-merge')
+                    return { state: s, result: undefined };
+                let n = (0, state_1.transitionBatch)(s, batch.id, 'merged', {}, now);
+                n = (0, state_1.transitionBatch)(n, batch.id, 'deployed', {}, now);
+                for (const issue of b.members) {
+                    const entry = (0, state_1.findEntry)(n, issue);
+                    if (entry && entry.status !== 'shipped-in-batch') {
+                        try {
+                            n = (0, state_1.transitionIssue)(n, issue, 'shipped-in-batch', {}, now);
+                            n = (0, state_1.transitionIssue)(n, issue, 'done', {}, now);
+                        }
+                        catch {
+                            // Already terminal via another rail — leave it.
+                        }
+                    }
+                }
+                return { state: n, result: undefined };
+            });
+            result.mergeAccepted.push(unit(batch.id));
+            continue;
+        }
+        if (truth.blocked || truth.mergeable === 'CONFLICTING') {
+            journalEvent(deps, 'pr-watch-failed', unit(batch.id), {
+                reason: truth.blocked ? 'auto-merge-blocked' : 'pr-conflicting',
+                pr,
+            });
+            // #472's own rebase-and-reship path (RFC F.9) is a documented follow-up
+            // for the batch PR-conflict rail; for now the batch stays parked and
+            // the block is visible via the journal + `sched status`.
+        }
+    }
+}
+/**
+ * Remove the batch's shared worktree — called both on the happy path
+ * (`reconcileReportSlot`, after `batch-report done`) and on every dissolve
+ * path (`runValidate`'s unattributable-suite dissolve, `evictMemberDirectly`'s
+ * eviction-threshold dissolve): a dissolved batch's worktree is otherwise
+ * left on disk forever, since nothing else ever calls this for it.
+ */
+function teardownBatch(deps, batchId) {
+    const state = deps.store.load();
+    const batch = (0, state_1.findBatch)(state, batchId);
+    if (!batch || batch.worktree === null)
+        return;
+    const root = deps.exec('git', ['rev-parse', '--show-toplevel'], deps.repoDir) ?? deps.repoDir;
+    if (!(0, teardown_1.isSafeWorktree)(path.resolve(root), batch.worktree)) {
+        journalEvent(deps, 'teardown-failed', unit(batchId), {
+            reason: 'unsafe-worktree-path',
+            detail: batch.worktree,
+        });
+        return;
+    }
+    const result = (0, teardown_1.runTeardown)(deps.exec, deps.repoDir, { worktree: batch.worktree, poolClaimed: false, branch: batch.branch }, deps.fsExists);
+    journalEvent(deps, result.cleanup === 'done' ? 'teardown-done' : 'teardown-failed', unit(batchId), {
+        cleanup: result.cleanup,
+        detail: result.detail,
+        worktree: batch.worktree,
+    });
+}
+// --- Entry point ---
+/**
+ * One batch reconcile+refill pass. Called from `engine.ts`'s `tick()` after
+ * the issue-level pass — batches never compete with issues for a slot within
+ * the same tick because this pass runs strictly after `dispatchAssignments`
+ * already filled every slot it could (see the module doc: batch claims never
+ * go through `computeAssignments`/`runnableUnits` at all). Loads and saves
+ * state itself via `deps.store.withLock` — the caller holds no lock across
+ * this call. `deps.exec` and `deps.runSuite` are mandatory; `deps.
+ * runCapability` is independently optional (AC2's incremental gate is itself
+ * a "when available" fast path).
+ */
+function runBatchTick(deps, config, dispatch) {
+    const result = emptyResult();
+    const now = deps.now();
+    for (const batch of deps.store.load().batches) {
+        if (batch.status === 'ready' && slotFor(deps.store.load(), batch.id) === undefined) {
+            claimAndSetup(deps, config, dispatch, batch.id, now, result);
+        }
+    }
+    for (const batch of deps.store.load().batches) {
+        const slot = slotFor(deps.store.load(), batch.id);
+        if (slot && (slot.status === 'running' || slot.status === 'assigned')) {
+            if (batch.status === 'executing') {
+                reconcileMemberSlot(deps, config, dispatch, batch.id, slot, now, result);
+            }
+            else if (batch.status === 'fixing') {
+                reconcileFixSlot(deps, config, batch.id, slot, now, result);
+            }
+            else if (batch.status === 'reviewing' || batch.status === 'shipping') {
+                reconcileTailSlot(deps, batch.id, slot, now, result);
+            }
+            else if (batch.status === 'deployed') {
+                reconcileReportSlot(deps, batch.id, slot, now, result);
+            }
+            continue;
+        }
+        if (slot)
+            continue; // live but neither running/assigned (e.g. mid-verify) — next tick
+        if (batch.status === 'validating') {
+            runValidate(deps, config, dispatch, batch.id, now, result);
+        }
+        else if (batch.status === 'deployed') {
+            spawnReportAgent(deps, config, dispatch, batch.id, now, result);
+        }
+        else if (batch.status === 'executing') {
+            // A prior spawn threw, or `claimAndSpawn` found zero free capacity —
+            // either way the batch is stuck mid-member with no slot and nothing
+            // else will ever retry it (Conformance review AC5 caveat; Supportability
+            // review #12). Retrying every tick is safe: `claimAndSpawn` itself is
+            // the capacity gate, so this is a no-op until a slot actually frees up.
+            spawnMemberContinuation(deps, config, dispatch, batch.id, now, result);
+        }
+        else if (batch.status === 'reviewing' || batch.status === 'shipping') {
+            // Same wedge, for a dead-or-never-claimed tail agent.
+            spawnTailAgent(deps, config, dispatch, batch.id, now, result);
+        }
+        else if (batch.status === 'fixing') {
+            // `beginFixAttempt` already recorded this member's ONE attempt as
+            // `dispatched` before `claimAndSpawn` could find capacity — retrying the
+            // exact same dispatch isn't reconstructible from persisted state (only
+            // the outcome is persisted, not the command/prompt), so resolve it
+            // `red` (conservatively: the member loses its one attempt and evicts on
+            // the next validate pass, which is safe — never a permanent wedge).
+            const state = deps.store.load();
+            const b = (0, state_1.findBatch)(state, batch.id);
+            const offenderRecord = b
+                ? [...b.fix_attempts].reverse().find((a) => a.outcome === 'dispatched')
+                : undefined;
+            if (b && offenderRecord) {
+                const rDeps = recoveryDeps(deps, b, now);
+                const { state: resolved } = (0, recovery_1.resolveFixAttempt)(state, batch.id, offenderRecord.issue, 'red', rDeps);
+                deps.store.withLock((s) => ({
+                    state: applyBatchAndIssues(s, resolved, batch.id, []),
+                    result: undefined,
+                }));
+            }
+        }
+    }
+    reconcilePrWatch(deps, now, result);
+    return result;
+}
+//# sourceMappingURL=batch-dispatch.js.map