tickmarkr 2.6.2 → 2.6.3

package/dist/run/daemon.js

@@ -1,5 +1,5 @@
 import { HOST_PROBE_SAMPLE_MS, HOST_PROBE_SAMPLES, HostDegradedError, hostDegraded, observeHost } from "./host-health.js";
-import { VITEST_CACHE_ENV, worktreeVitestCache } from "../gates/test-manifest.js";
+import { runWithInterruptionSink, VITEST_CACHE_ENV, worktreeVitestCache } from "../gates/test-manifest.js";
 import { COMMAND_LEASE_TOKEN_ENV, commandLeaseEnvironment, CommandLeases, currentCommandLeaseToken, isRunnerCommand, runWithCommandLease, withCommandLease } from "./lease.js";
 import { execFileSync, spawn } from "node:child_process";
 import { createHash, randomBytes } from "node:crypto";
@@ -21,7 +21,7 @@ import { driverEvidence } from "../drivers/index.js";
 import { herdrSealShellPrefix, MAX_BUF, SubprocessDriver } from "../drivers/subprocess.js";
 import { formatOwnedName } from "../drivers/types.js";
 import { captureBaseline, detectGateCommands, detectVacuousOracles } from "../gates/baseline.js";
-import { runGates } from "../gates/run-gates.js";
+import { reobserveTestFiles, runGates } from "../gates/run-gates.js";
 import { isInfraResult } from "../gates/cache.js";
 import { filesGlob } from "../graph/files-glob.js";
 import { addEvidence, attributeBlocked, batteryPriority, blockedTasks, getTask, graphDefinitionHash, loadGraph, pendingTasks, readyTasks, saveGraph, setStatus, taskContentDigest, tickmarkrDir } from "../graph/graph.js";
@@ -29,17 +29,17 @@ import { GATE_NAMES } from "../graph/schema.js";
 import { distFingerprint } from "../cli/commands/version.js";
 import { augmentRetryBrief, consult, renderRetryGuidance } from "./consult.js";
 import { executionSignal, remainingExecutionMs, withExecutionBudget, withoutExecutionBudget, ExecutionBudgetExceeded } from "./execution-budget.js";
-import { repairSelectionDecision } from "./repair-selection.js";
+import { repairSelectionDecision, repairSelectionEnabled } from "./repair-selection.js";
 import { failureDisposition, reserveInfrastructureRetry } from "./recovery.js";
 import { runEnvironment } from "./environment.js";
 import { changeRepresented, cleanupRunWorktrees, deriveForkCap, FORK_CAP_ENV, gitHead, linkNodeModules, npmDependencyInstallCommand, npmDependencyManifestChanged, PRESERVE_COMMIT_SUBJECT, PRESERVE_PRODUCER_TRAILER, preserveWorktree, producerFields, resolvedCapacity, runWithForkBudget, runWithVerificationBudget, sameCapacity, sameVerification, sh, shGit, SUITE_PARENT_ENV, verificationProtocol, WORKTREE_LAYOUT_CONTRACT, worktreePath } from "./git.js";
 import { runInteractiveSeed } from "./interactive-seed.js";
 import { classifyRepairDisposition, resolveScopeHints } from "./repair-disposition.js";
-import { applyScopeAmendments, activeRetryBan, interruptedAttempt, RESUME_HARVEST_SOURCE, resumeHarvestAuthor, approvalAction, APPROVAL_REFUSED, bindingToken, effectiveEvents, foldDecisions, physicalLine, staleApprovals, classifyTaskFailure, classifyWorkerResultCause, deferredReviewFindings, engagementComparable, formatPriorFindingEvidence, GATE_FINGERPRINT_CAP, GATE_SATISFIED_RELEASE, identicalGateFailures, isDeferredFinding, journaledFailureBrief, Journal, loadRoutingProfile, newRunId, normalizeGateFailure, outstandingConsultGuidance, outstandingReviewFindings, pendingApprovalActions, pendingRechecks, pendingRepairFindings, phaseForGate, readPriorRunEvidence, recordedTaskFailureKind, RECHECK_RELEASE, renderStructuredReviewFinding, repairReachSinceApproval, repairsSinceApproval, reviewRoundsSinceApproval, runHasEnded, standingRulings, structuredFindings, upheldFeedbackByTask } from "./journal.js";
-import { gateReviewerFloor, isDiffCapPark, pickReviewer } from "../gates/review.js";
+import { applyScopeAmendments, activeRetryBan, interruptedAttempt, RESUME_HARVEST_SOURCE, resumeHarvestAuthor, approvalAction, APPROVAL_REFUSED, bindingToken, effectiveEvents, foldDecisions, physicalLine, staleApprovals, classifyTaskFailure, classifyWorkerResultCause, deferredReviewFindings, engagementComparable, formatPriorFindingEvidence, GATE_FINGERPRINT_CAP, GATE_SATISFIED_RELEASE, identicalGateFailures, isDeferredFinding, journaledFailureBrief, Journal, loadRoutingProfile, newRunId, normalizeGateFailure, outstandingConsultGuidance, readableExcerpt, receiptOrigin, outstandingReviewFindings, pendingApprovalActions, pendingRechecks, pendingRepairFindings, phaseForGate, priorJudgments, readPriorRunEvidence, recordedTaskFailureKind, reraisedReviewChains, RECHECK_RELEASE, renderStructuredReviewFinding, repairReachSinceApproval, repairsSinceApproval, reviewRoundsSinceApproval, runHasEnded, standingRulings, structuredFindings, upheldFeedbackByTask } from "./journal.js";
+import { gateReviewerFloor, isDiffCapPark, isGarbageReview, pickReviewer } from "../gates/review.js";
 import { acquireApprovalSerialization, acquireRunLock, isPidLive, releaseRunLock } from "./lock.js";
 import { ensureIntegration, integrationBranch, integrationHead, mergeTask, reusedTipEvidence, verifyIntegrationTip } from "./merge.js";
-import { climbChannel, marginalCostRank, nextChannel, route } from "../route/router.js";
+import { climbChannel, inTaskPool, marginalCostRank, nextChannel, poolExhaustion, route, taskPool } from "../route/router.js";
 import { desiredPanes } from "./reconcile.js";
 import { readTierLiveness, readWatchBoard, supervisionPresencePath } from "./supervision.js";
 import { harvestCpuFlatWindowMs, NUDGEABLE_ADAPTERS, PANE_READ_ROWS, StallProgressTracker, stallSnapshotBannerRows, WorkerTreeCpuAccountant, reapOwnedProcessGroup, readOwnedProcessGroup, } from "./stall.js";
@@ -370,6 +370,30 @@ function classifyInfraResult(g) {
         return;
     g.meta = { ...g.meta, classification: "infra", infra: true };
 }
+// OBS-1106 residual: a red whose assertion is a timeout or a wall-clock budget, and the runner-level
+// diagnostics (never-started files, a worker RPC timeout) that make such a red infrastructure-SHAPED.
+// Shape alone never parks anything: it only buys one isolated re-observation before a charge. A
+// fresh-fingerprint section masks digits to `#`, so both spellings of a number are read.
+const TIMEOUT_SHAPED_RE = /\b(?:Test|Hook) timed out in (?:\d+|#)\s*ms\b|\bexpected (?:\d+|#)(?:\.(?:\d+|#))? to be (?:less than|below)(?: or equal to)? (?:\d+|#)/i;
+const RUNNER_INFRA_DIAGNOSTIC_RE = /\bnever-started (?:[1-9]\d*|#)(?![\d#])|\[vitest-worker\]: Timeout calling\b/;
+const timeoutShaped = (details) => TIMEOUT_SHAPED_RE.test(readableExcerpt(details));
+const ATTRIBUTED_FILE_RE = /^\s*(?:FAIL|×|✗)\s+(?:\|[^|\n]*\|\s+)?(\S+\.(?:test|spec)\.[cm]?[jt]sx?)\b/gm;
+/** The failing files a red attributes to itself: the runner's own list, else its FAIL headlines. */
+const attributedFailingFiles = (g) => [...new Set(Array.isArray(g.meta?.failingFiles) && g.meta.failingFiles.length
+        ? g.meta.failingFiles.filter((f) => typeof f === "string")
+        : [...readableExcerpt(g.details).matchAll(ATTRIBUTED_FILE_RE)].map((m) => m[1].replace(/^\.\//, "")))].sort();
+const metaFiles = (v) => Array.isArray(v) ? v.filter((f) => typeof f === "string") : undefined;
+/** D-607: the files a rerun OBSERVABLY ran — the runner's manifest, the only complete selection
+ * evidence. Never the requested selection, and never the per-file lines a runner printed: a runner that
+ * ignores file arguments and reports only its failures looks isolated while it ran the whole suite.
+ * Undefined without a manifest, and the rerun is then ambiguous. */
+const observedSelection = (g) => {
+    const ran = metaFiles(g.meta?.manifest);
+    return ran?.length ? [...new Set(ran)].sort() : undefined;
+};
+/** D-607: on the manifest path only the reporter's own attribution counts — a `FAIL path` line inside
+ * captured stdout is text, not attribution. Headline lexing is for runners without a reporter. */
+const rerunFailingFiles = (g) => metaFiles(g.meta?.manifest) ? [...new Set(metaFiles(g.meta?.failingFiles) ?? [])].sort() : attributedFailingFiles(g);
 // v1.85 T3: the gates whose failure IS a deterministic measurement — a machine re-ran a command over a
 // tree and printed the same bytes. Those are the failures the fingerprint cap governs (the ruling names
 // it a "deterministic-gate" cap): a third identical answer to a question already answered twice is the
@@ -460,18 +484,36 @@ function repairDiffFilesCut(diff, cap, names) {
 function repairFindingsBrief(results) {
     return results.filter(gateFailed).map((g) => {
         const evidence = g.meta?.failureEvidence;
-        return `${g.gate}: ${g.details}` + (Array.isArray(evidence) && evidence.length
+        return `${g.gate}: ${readableExcerpt(g.details)}` + (Array.isArray(evidence) && evidence.length
             ? `\nAssertion evidence:\n${JSON.stringify(evidence, null, 2)}` : "");
     }).join("\n\n");
 }
-/** The fix-only contract: the findings verbatim, then the diff content of the work already landed. */
-function repairBrief(findings, diff, baseRef) {
+// OBS-1019 add.2: the order a repeated review chain earns in place of the smallest-edit order.
+const REPAIR_CLASS_SENTENCE = "Before editing a chain reraised more than once, enumerate the in-scope consumers,"
+    + " bridges, operations and event sequences as a closed case table; repair every implicated member and verify"
+    + " the accumulated diff against that table.";
+/**
+ * The fix-only contract: the findings verbatim, then the diff content of the work already landed.
+ * OBS-1019 add.2: a first finding stays bounded to the smallest edit. A chain a review has re-raised
+ * more than once was chased edge by edge while every round re-judged the accumulated diff, so it owes
+ * the whole class instead: a closed case table inside the declared bounds, never beyond them.
+ */
+function repairBrief(findings, diff, baseRef, repeated = []) {
     const repair = { findings, diff };
     return [
-        "## Repair attempt — fix ONLY what these findings name",
+        repeated.length ? "## Repair attempt — repair the whole class these findings name" : "## Repair attempt — fix ONLY what these findings name",
         "The commits from your prior attempt are already in this worktree and their diff is reproduced"
-            + " below. Do NOT re-implement that work, do not start over, and do not revert it: make the"
-            + " smallest change that resolves every finding, then commit.",
+            + " below. Do NOT re-implement that work, do not start over, and do not revert it: "
+            + (repeated.length ? "resolve every member of the case table below, then commit." : "make the smallest change that resolves every finding, then commit."),
+        ...(repeated.length ? [
+            "",
+            "### Repeated review chains — repair the class, not the next edge",
+            // Fingerprint and count only: the finding's prose is already quoted once, below or under its heading.
+            ...repeated.map(({ finding, reraises }) => `- \`${finding.fingerprint}\` re-raised ${reraises} times`),
+            REPAIR_CLASS_SENTENCE,
+            "The table stays inside the task's declared write scope (files[]) and never narrows its goal or an"
+                + " acceptance criterion; the reviewer judges the accumulated diff, not the table.",
+        ] : []),
         "",
         "### Failing gate findings (verbatim)",
         repair.findings,
@@ -881,7 +923,7 @@ export function formatTipProof(p) {
 }
 // Isolate the battery so cancellation can stop its control flow as well as its shell children.
 // The child uses the same verifier and capacity; only the daemon writes lifecycle verdict rows.
-async function cancellableTipBattery(intWt, commands, runDir, baseline, signal) {
+async function cancellableTipBattery(intWt, commands, runDir, baseline, signal, evidence = {}) {
     signal.throwIfAborted();
     const extension = import.meta.url.endsWith(".ts") ? "ts" : "js";
     const script = `
@@ -930,9 +972,9 @@ async function cancellableTipBattery(intWt, commands, runDir, baseline, signal)
       finally { process.send({ type: 'lease-release', id }); }
     };
     let input = ''; for await (const chunk of process.stdin) input += chunk;
-    const { intWt, commands, runDir, baseline, capacity } = JSON.parse(input);
+    const { intWt, commands, runDir, baseline, capacity, evidence } = JSON.parse(input);
     try {
-      const result = await runWithVerificationBudget(capacity, () => runWithCommandLease(lease, () => verifyIntegrationTip(intWt, commands, runDir, baseline)));
+      const result = await runWithVerificationBudget(capacity, () => runWithCommandLease(lease, () => verifyIntegrationTip(intWt, commands, runDir, baseline, evidence)));
       if (!cancelled) process.stdout.write(JSON.stringify(result.map(row => ({ ...row, capacity: [...commandCapacities].reverse().find(([command]) => command === row.cmd || command.startsWith(row.cmd + " "))?.[1] ?? capacity }))));
     } catch (error) { if (!cancelled) throw error; }
     finally { process.disconnect(); }
@@ -983,7 +1025,7 @@ async function cancellableTipBattery(intWt, commands, runDir, baseline, signal)
     });
     signal.addEventListener("abort", cancel, { once: true });
     child.stdin.on("error", () => { });
-    child.stdin.end(JSON.stringify({ intWt, commands, runDir, baseline, capacity: resolvedCapacity() }));
+    child.stdin.end(JSON.stringify({ intWt, commands, runDir, baseline, capacity: resolvedCapacity(), evidence }));
     try {
         if (signal.aborted)
             cancel();
@@ -1044,9 +1086,9 @@ export async function verifyIntegrationTipCached(intWt, commands, journal, opts
         return false;
     }
     let tipFailed = false;
-    for (const r of await (opts.signal
-        ? cancellableTipBattery(intWt, commands, journal.dir, opts.baseline, opts.signal)
-        : verifyIntegrationTip(intWt, commands, journal.dir, opts.baseline))) {
+    for (const r of await runWithInterruptionSink(journalInterruption(journal, undefined), () => opts.signal
+        ? cancellableTipBattery(intWt, commands, journal.dir, opts.baseline, opts.signal, opts.evidence)
+        : verifyIntegrationTip(intWt, commands, journal.dir, opts.baseline, opts.evidence))) {
         const measuredCapacity = r.capacity ?? capacity;
         const evidence = r.reused ? reusedTipEvidence(r) : {
             ...(r.originRunRoot ? { originRunRoot: r.originRunRoot } : {}),
@@ -1080,6 +1122,12 @@ export async function verifyIntegrationTipCached(intWt, commands, journal, opts
     }
     return tipFailed;
 }
+/** OBS-953: a detected wall-over-monotonic discontinuity journals `host-suspend` (its offset was
+ * subtracted from active test time); an overdue poll whose clocks advanced together journals
+ * `host-poll-overdue` (recorded, never subtracted). */
+const journalInterruption = (journal, taskId) => (interruption) => {
+    journal.append(interruption.kind === "host-suspend" ? "host-suspend" : "host-poll-overdue", taskId, { ...interruption });
+};
 async function commitsAheadOfRef(base, head, cwd) {
     if (head === base)
         return [];
@@ -1904,6 +1952,9 @@ export async function runDaemon(repoRoot, opts = {}) {
         // > repo > global > default. The resolved cfg carries mode-compiled floors; route() never sees the mode.
         const rm = resolveRunMode(repoRoot, { flag: opts.mode, spec: graph.mode, globalDir: opts.globalDir });
         const cfg = rm.cfg;
+        // OBS-1199: the effective repair-selection default is resolved apart from the optional execution
+        // budget, so an absent policy selects on every driver; a resume keeps what the run recorded.
+        let repairSelection = repairSelectionEnabled(cfg);
         if (opts.resume) {
             const start = Journal.open(repoRoot, runId).read().find((e) => e.event === "run-start");
             const snapshot = start?.data.effectivePolicy;
@@ -1915,6 +1966,11 @@ export async function runDaemon(repoRoot, opts = {}) {
             // Historical runs did not opt in; current settings cannot grant them a new allowance.
             cfg.executionPolicy = recorded?.config?.executionPolicy === undefined
                 ? undefined : ExecutionPolicySchema.parse(recorded.config.executionPolicy);
+            const recordedGates = recorded?.config?.gates;
+            if (recordedGates?.repairSelection !== undefined && typeof recordedGates.repairSelection !== "boolean") {
+                throw new Error("cannot restore malformed recorded repair selection");
+            }
+            repairSelection = repairSelectionEnabled({ executionPolicy: cfg.executionPolicy, gates: recordedGates });
         }
         if (cfg.executionPolicy && driver.id !== "subprocess") {
             throw new Error("experimental executionPolicy currently requires the subprocess driver");
@@ -2366,6 +2422,19 @@ export async function runDaemon(repoRoot, opts = {}) {
                     void promise.finally(() => hostChecks.delete(promise)).catch(() => { });
                     return promise;
                 };
+                // OBS-1190: every admission poll still probes (the probe gates admission), but the journal keeps only
+                // the first observation and each change of state or reference — healthy, degraded, unreadable,
+                // recovered — the way suite-wait already dedupes by count. A steady wait no longer floods the journal.
+                let journaledHost;
+                const journalObservation = (taskId, observation, resuming) => {
+                    const state = observation.medianMs === null ? "unreadable"
+                        : hostReferenceMs !== undefined && hostDegraded(observation, hostReferenceMs) ? "degraded" : "healthy";
+                    const key = `${state}:${hostReferenceMs ?? ""}`;
+                    if (key === journaledHost)
+                        return;
+                    journaledHost = key;
+                    journal.append("host-observation", taskId, { ...observation, referenceMs: hostReferenceMs ?? null, resuming, state });
+                };
                 const recordReference = (observation) => {
                     if (observation.medianMs === null)
                         return;
@@ -2389,7 +2458,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                         const probeBudget = HOST_PROBE_SAMPLE_MS * HOST_PROBE_SAMPLES;
                         if (!observation || (!first && remaining >= probeBudget)) {
                             observation = await observeHost(signal, remaining > 0 ? Math.min(probeBudget, remaining) : undefined);
-                            journal.append("host-observation", taskId, { ...observation, referenceMs: hostReferenceMs ?? null, resuming });
+                            journalObservation(taskId, observation, resuming);
                         }
                         first = false;
                         if (hostReferenceMs === undefined && observation.medianMs !== null)
@@ -2433,7 +2502,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                 const initializeHost = () => trackHost((async () => {
                     const signal = hostSignal();
                     const observation = await observeHost(signal);
-                    journal.append("host-observation", undefined, { ...observation, referenceMs: hostReferenceMs ?? null, resuming: !!opts.resume });
+                    journalObservation(undefined, observation, !!opts.resume);
                     if (hostReferenceMs === undefined)
                         recordReference(observation);
                     if (opts.resume || observation.medianMs === null)
@@ -2479,7 +2548,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                             waited = true;
                             journal.append("suite-wait", taskId, { count, ...(gate ? { gate } : {}) });
                         }, SUITE_POLL_MS, signal);
-                    }, async () => {
+                    }, () => runWithInterruptionSink(journalInterruption(journal, taskId), async () => {
                         try {
                             const result = await run();
                             // Some command oracles turn launch errors into results. Admission failure still parks infra.
@@ -2491,7 +2560,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                             if (taskId)
                                 activeGatePhases.delete(taskId);
                         }
-                    });
+                    }));
                 };
                 let baseRef;
                 let baseline;
@@ -2710,24 +2779,39 @@ export async function runDaemon(repoRoot, opts = {}) {
                 // owns listing/parsing/closing. Cosmetic by contract: failures are swallowed and subprocess has no
                 // reconcile (optional chain → no-op), so gates and the oracle suite never feel this. keepPanes
                 // "forever" is the keep-everything debug override — it disables the sweep entirely.
-                const resuming = !!opts.resume;
+                // OBS-1109: an interrupted attempt's pane is evidence — its nonce-bound trailer decides whether
+                // the attempt is harvested, declined as foreign, or recovered — so the fold's run-resume clear
+                // must not reach it before harvestInterruptedAttempt has read it and reaped its processes. Only a
+                // resumed daemon can hold such an attempt: a fresh run's sweep is the baseline sweep.
+                // OBS-1204: the set is FROZEN here, once, from the rows up to this daemon's own run-resume row: an
+                // attempt launched after it is this daemon's own and reconciles exactly as in a fresh run. Each
+                // frozen spare lapses for good once its harvest, decline or a superseding dispatch is journaled.
+                const resumeInterrupted = new Map();
+                if (opts.resume) {
+                    const rows = journal.read();
+                    let resumedAt = rows.length - 1;
+                    while (resumedAt >= 0 && rows[resumedAt].event !== "run-resume")
+                        resumedAt--;
+                    const prior = rows.slice(0, resumedAt + 1);
+                    for (const t of graph.tasks) {
+                        const owned = interruptedAttempt(prior, t.id);
+                        if (owned?.launch)
+                            resumeInterrupted.set(t.id, { attempt: owned.attempt, nonce: owned.launch.nonce });
+                    }
+                }
                 const reconcile = async (sweep) => {
                     if (keepForever)
                         return;
                     try {
                         const rows = journal.read();
                         const desired = desiredPanes(rows, runId);
-                        // OBS-1109: an interrupted attempt's pane is evidence — its nonce-bound trailer decides whether
-                        // the attempt is harvested, declined as foreign, or recovered — so the fold's run-resume clear
-                        // must not reach it before harvestInterruptedAttempt has read it and reaped its processes. Once
-                        // that harvest (or a superseding dispatch) is journaled the fold no longer names it: swept then.
-                        // Only a resumed daemon can hold such an attempt: a fresh run's sweep is the baseline sweep.
-                        if (resuming)
-                            for (const t of graph.tasks) {
-                                const owned = interruptedAttempt(rows, t.id);
-                                if (owned?.launch)
-                                    desired.add(formatOwnedName({ role: "worker", taskId: t.id, attempt: owned.attempt, runId }));
-                            }
+                        for (const [taskId, frozen] of resumeInterrupted) {
+                            const owned = interruptedAttempt(rows, taskId);
+                            if (owned?.launch?.nonce !== frozen.nonce)
+                                resumeInterrupted.delete(taskId);
+                            else
+                                desired.add(formatOwnedName({ role: "worker", taskId, attempt: frozen.attempt, runId }));
+                        }
                         // The watch pane is never the DRIVER sweep's candidate (panesToClose spares role "watch":
                         // herdr's watches bookkeeping lives in close(), and a raw pane-close in the sweep would
                         // leave narrator() a stale cache) — the driver always sees it as desired; its lifecycle is
@@ -2835,6 +2919,80 @@ export async function runDaemon(repoRoot, opts = {}) {
                     await reconcile({ spareLiveLlm: true }); // task-human is a terminal event — sweep, sparing sibling tasks' live LLM panes
                     await driver.notify(`tickmarkr ${runId}: ${t.id} needs a human — ${reason}${token ? `\npark ${token} — bind the decision with \`--park ${token}\`` : ""}`, { tier: "attention" });
                 };
+                // OBS-1106 residual: a retry that landed nothing on a timeout-class red re-observes it ONCE fresh:
+                // neither a journal replay nor the verdict cache may re-state that red. Owed when the prior round's
+                // test red sits on this very subject, is timeout-shaped, and has not yet been observed twice
+                // independently — once it has, the fingerprint cap owns the next move. Read from the journal alone,
+                // so the live gate loop and a resume battery decide alike.
+                const owesTimeoutRefresh = (t, priorRows, commit) => {
+                    const red = priorRows.find((e) => e.event === "gate-result" && e.data.gate === "test" && e.data.pass === false
+                        && e.data.skipped !== true && e.data.commit === commit && typeof e.data.details === "string" && timeoutShaped(e.data.details));
+                    return red !== undefined
+                        && identicalGateFailures(journal.read(), t.id, "test", normalizeGateFailure(String(red.data.details))) < GATE_FINGERPRINT_CAP;
+                };
+                // OBS-1106 residual: before a timeout or wall-budget red accompanied by runner infra diagnostics is
+                // charged, its attributed failing files are re-run ONCE, on the same checkout, under the suite
+                // lease. Only a reproduced assertion is charged. A pass, an infra rerun, a rerun naming none of the
+                // files, or an unattributable red parks with the ORIGINAL red journaled and unresolved: the rerun is
+                // diagnostic, never a verdict, so it can neither green the gate nor permit a merge. Returns true
+                // when it parked.
+                const adjudicateInfraShapedRed = async (t, results, wt, subject, attempt, replayed, parkInfra) => {
+                    const red = results.find((g) => g.gate === "test" && gateFailed(g) && !isInfraResult(g)
+                        && timeoutShaped(g.details) && RUNNER_INFRA_DIAGNOSTIC_RE.test(g.details));
+                    if (!red || commands.test === undefined)
+                        return false;
+                    // The observation this red states: its subject, its fingerprint and, when it carries one, its
+                    // receipt's invocation. A copy (journal replay, cache hit) is never re-run when THAT observation's
+                    // adjudication is on the ledger — its recorded outcome decides, so a crash after the adjudication
+                    // but before its park still parks. A crash before the adjudication leaves none, so the copy is
+                    // adjudicated now. A fresh red always is.
+                    const fingerprint = normalizeGateFailure(red.details);
+                    const observation = receiptOrigin({ evidenceReceipt: red.evidenceReceipt ?? red.meta?.evidenceReceipt,
+                        evidenceReceipts: red.evidenceReceipts ?? red.meta?.evidenceReceipts });
+                    const recorded = replayed || red.meta?.reused === true
+                        ? journal.read().reverse().find((e) => e.taskId === t.id && e.event === "gate-reobserved"
+                            && e.data.commit === subject && e.data.fingerprint === fingerprint
+                            && (observation === undefined || e.data.observation === undefined || e.data.observation === observation))
+                        : undefined;
+                    let files = [];
+                    let outcome = "unattributed";
+                    let selection;
+                    if (recorded) {
+                        files = Array.isArray(recorded.data.files) ? recorded.data.files.filter((f) => typeof f === "string") : [];
+                        outcome = ["reproduced", "passed", "unattributed"].includes(String(recorded.data.outcome)) ? recorded.data.outcome : "ambiguous";
+                    }
+                    else {
+                        for (const file of attributedFailingFiles(red)) {
+                            if (posix.normalize(file) === file && !file.startsWith("../") && !file.startsWith("/") && existsSync(join(wt, file)))
+                                files.push(file);
+                        }
+                        let rerun;
+                        if (files.length) {
+                            rerun = await withCommandContext(t.id, () => reobserveTestFiles(wt, commands.test, baseline, files, journal.dir))
+                                .catch((error) => ({ gate: "test", pass: false, details: `infra: rerun failed: ${error instanceof Error ? error.message : String(error)}`, meta: { infra: true } }));
+                            classifyInfraResult(rerun);
+                            // Reproduced only by failure evidence the rerun ATTRIBUTES to one of exactly these files, from an
+                            // OBSERVED selection that was exactly these files (D-607): a path merely printed in output, a
+                            // positional filter that also collected another file, or a runner that ignored the file
+                            // arguments and reported no manifest of its selection, says nothing about the original red.
+                            const failing = rerunFailingFiles(rerun);
+                            selection = observedSelection(rerun);
+                            const isolated = selection !== undefined && selection.length === files.length
+                                && files.every((file) => selection.includes(file)) && failing.every((file) => files.includes(file));
+                            outcome = rerun.pass ? "passed" : isolated && !isInfraResult(rerun) && failing.length ? "reproduced" : "ambiguous";
+                        }
+                        journal.append("gate-reobserved", t.id, {
+                            gate: "test", attempt, commit: subject, fingerprint, ...(observation ? { observation } : {}), files, outcome,
+                            ...(selection ? { selection } : {}),
+                            ...(rerun ? { pass: rerun.pass, details: rerun.details } : {}),
+                            ...(rerun?.evidenceReceipt ? { evidenceReceipt: rerun.evidenceReceipt } : {}),
+                        });
+                    }
+                    if (outcome === "reproduced")
+                        return false;
+                    await parkInfra(`test: timeout-shaped red with runner infra diagnostics was not reproduced by one isolated rerun of ${files.length ? files.join(", ") : "its attributed files (none attributable)"} (${outcome}); the original red stands unresolved — recheck to re-measure:\n${readableExcerpt(red.details)}`);
+                    return true;
+                };
                 // OBS-547: cross-reference a scope red against the prediction this run already computed. Every hard
                 // offender predicted ⇒ an AUTHORING defect whose repair is pre-written: journal the classification
                 // with the verbatim files[] lines and park unchargeable. It owes nothing to the quality machinery —
@@ -2853,7 +3011,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                     const verdict = scopeRed?.meta?.collateral;
                     const worker = journal.read().reverse().find((e) => e.taskId === t.id && e.event === "worker-result");
                     const disposition = classifyRepairDisposition({
-                        results, files: t.files, inventory: changed,
+                        // OBS-1072 add.2: hints are lexed from the readable excerpt, never from terminal control bytes.
+                        results: results.map((g) => ({ ...g, details: readableExcerpt(g.details) })), files: t.files, inventory: changed,
                         refusalSummary: worker?.data.ok === false && typeof worker.data.summary === "string" ? worker.data.summary : undefined,
                     });
                     for (const diagnostic of disposition.diagnostics) {
@@ -2965,7 +3124,13 @@ export async function runDaemon(repoRoot, opts = {}) {
                     if (changed)
                         saveGraph(repoRoot, graph);
                 };
+                // OBS-1168(c): every gate callback is bound to the engagement that asked for it. Once the scheduler
+                // settles that engagement (done, parked or failed), a late sibling's event is dropped, never journaled.
+                const engagements = new Map();
                 const execTask = async (t) => {
+                    const engagement = {};
+                    engagements.set(t.id, engagement);
+                    const liveEngagement = () => engagements.get(t.id) === engagement;
                     const approvalAction = pendingDaemonApprovalActions(journal.read()).get(t.id);
                     const fundedRerun = approvalAction?.authority === "worker";
                     const startMs = Date.now();
@@ -3049,8 +3214,10 @@ export async function runDaemon(repoRoot, opts = {}) {
                     // OBS-1161: no `attempts > 0` guard — every release already clears lastAssignment in the replay,
                     // and a lastAssignment at zero attempts is a first dispatch whose capacity requeue was taken back:
                     // the seat is still in force, so restore it instead of failing over its own tried[] entry early.
+                    // OBS-1187: a restored seat must still be a member of the task's pool (the operator may have closed it).
                     if (!hintsChanged && rs?.lastAssignment
                         && channels.some((c) => channelKey(c) === channelKey(rs.lastAssignment))
+                        && inTaskPool(t, cfg, channelKey(rs.lastAssignment))
                         && !demotedChannels.has(channelKey(rs.lastAssignment))) {
                         assignment = rs.lastAssignment; // restore the consult-chosen assignment (bypasses route()'s static re-pick)
                     }
@@ -3079,7 +3246,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                     const lastApproval = taskHistory.map((e) => e.event).lastIndexOf("task-approved");
                     const priorClimb = taskHistory.slice(lastApproval + 1).reverse().find((e) => e.event === "tier-escalated");
                     if (!hintsChanged && priorClimb && taskHistory.indexOf(priorClimb) > taskHistory.map((e) => e.event).lastIndexOf("task-dispatch")) {
-                        const target = channels.find((c) => channelKey(c) === priorClimb.data.to && !demotedChannels.has(channelKey(c)));
+                        const target = channels.find((c) => channelKey(c) === priorClimb.data.to && !demotedChannels.has(channelKey(c))
+                            && inTaskPool(t, cfg, channelKey(c)));
                         if (target)
                             assignment = seatAssignment(target);
                     }
@@ -3122,7 +3290,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                             workerDispatchOrdinal: previousDispatch?.data.workerDispatchOrdinal ?? null,
                         });
                     if (bootstrapExhausted === channelKey(assignment)) { // a later tier-climb restore may have moved it off
-                        await park(t, `${bootstrapExhausted} failed at bootstrap and no eligible channel remains after resume`, "infra", assignment, rs?.attempts ?? 0, startMs, 0, 0, undefined, 0, "fresh", { cause: "bootstrap", channel: bootstrapExhausted });
+                        await park(t, `${bootstrapExhausted} failed at bootstrap and no eligible channel remains after resume${poolExhaustion(t, cfg)}`, "infra", assignment, rs?.attempts ?? 0, startMs, 0, 0, undefined, 0, "fresh", { cause: "bootstrap", channel: bootstrapExhausted });
                         return;
                     }
                     // Keep one live list: recovery retries must see exclusions added by onGate during the round.
@@ -3133,18 +3301,33 @@ export async function runDaemon(repoRoot, opts = {}) {
                             if (!badReviewers.includes(e.payload.reviewer))
                                 badReviewers.push(e.payload.reviewer);
                         }
+                        // OBS-1196: a seat whose same-subject re-emission delivered a valid verdict was a delivery flake,
+                        // not a bad reviewer — it stays seatable for this task's later rounds.
+                        if (e.name === "review-reemission" && e.payload.delivered === true && typeof e.payload.reviewer === "string") {
+                            const at = badReviewers.indexOf(e.payload.reviewer);
+                            if (at >= 0)
+                                badReviewers.splice(at, 1);
+                        }
                     };
                     // Keep the measured siblings, including a red judge, while replacing only a missing review.
                     // Every dispatched seat is excluded for this task, so the loop is bounded by the eligible pool.
                     const runReviewRecovery = async (task, ctx, allowSelection = true) => {
                         ctx.authorizeInfraRetry = cfg.executionPolicy?.boundedInfrastructure
                             ? (subject, cause) => reserveInfrastructureRetry(journal.read(), task.id, subject, journal.append.bind(journal), cause) : undefined;
-                        if (cfg.executionPolicy?.repairSelection && allowSelection) {
+                        if (repairSelection && allowSelection) {
                             const decision = repairSelectionDecision(journal.read(), task.id, true);
                             ctx.selectTests = decision.selectTests;
                             ctx.requiredRepairTests = decision.requiredFiles;
                             ctx.selectionReason = decision.reason;
                         }
+                        else if (repairSelection) {
+                            // A recheck runs the full suite; its recorded scope keeps a red there attributable.
+                            ctx.selectTests = false;
+                            ctx.selectionReason = "recheck-full-suite";
+                        }
+                        // OBS-1151: every round judges fresh; the journal's earlier judgments (this run and before a
+                        // resume) are only the subjects a reversed ruling is compared against.
+                        ctx.priorJudgments = priorJudgments(journal.read(), task.id);
                         // Resume and ordinary verification share this boundary. Count persisted rounds so a
                         // resumed daemon cannot reuse a previous round's identity within the same attempt.
                         ctx.buildReceiptIdentity = {
@@ -3247,9 +3430,9 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 ...(typeof g.meta.producer === "string" ? { producer: g.meta.producer } : {}),
                                 ...(typeof g.meta.producerAttempt === "number" ? { producerAttempt: g.meta.producerAttempt } : {}),
                             } : {}),
-                            ...(cfg.executionPolicy && !g.pass ? { disposition: failureDisposition(g) } : {}),
+                            ...((cfg.executionPolicy || repairSelection) && !g.pass ? { disposition: failureDisposition(g) } : {}),
                             ...Object.fromEntries(["runnerInfraRerun", "hostStarvedRerun", "recoveryBlocked", "failingFiles", "selectionDecision", "failureEvidence",
-                                "forgivenFingerprints", "freshFingerprints", "baselineProvenance"]
+                                "forgivenFingerprints", "freshFingerprints", "baselineProvenance", "judgment", "adjudication", "judgeDisagreement", "interruptions"]
                                 .filter((key) => g.meta?.[key] !== undefined).map((key) => [key, g.meta[key]])),
                             // OBS-540: preserve terminal-vs-retryable infra exactly. normalizeGateOutcome deliberately
                             // defaults a legacy infra row to retryable, so dropping an explicit false here reverses the
@@ -3275,7 +3458,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                             } : {}),
                             ...(g.gate === "review" && typeof g.meta?.reviewer === "string" ? {
                                 reviewer: g.meta.reviewer,
-                                ...Object.fromEntries(["cause", "seatAuthoredBytes", "bytes", "rawPath", "briefPath", "timeoutMs", "unparseable", "noVerdict", "resolved", "reraised", "reviewerFloor", "reviewerFloorCause", "reviewerTier"]
+                                ...Object.fromEntries(["cause", "seatAuthoredBytes", "bytes", "rawPath", "briefPath", "timeoutMs", "unparseable", "noVerdict", "resolved", "reraised", "settledAnchors", "reviewerFloor", "reviewerFloorCause", "reviewerTier"]
                                     .filter((key) => g.meta?.[key] !== undefined).map((key) => [key, g.meta[key]])),
                                 ...(typeof g.meta.vendor === "string" ? { vendor: g.meta.vendor } : {}),
                                 ...(typeof g.meta.provider === "string" ? { provider: g.meta.provider } : {}),
@@ -3409,7 +3592,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                             return;
                         const from = channelKey(assignment);
                         const to = channelKey(next);
-                        const poolBefore = channels.filter((c) => !tried.includes(channelKey(c)) && !demotedChannels.has(channelKey(c))).map(channelKey);
+                        const poolBefore = channels.filter((c) => !tried.includes(channelKey(c)) && !demotedChannels.has(channelKey(c))
+                            && inTaskPool(t, cfg, channelKey(c))).map(channelKey);
                         journal.append("tier-escalated", t.id, {
                             attempt, cause, gate: gate.gate, fingerprint: normalizeGateFailure(gate.details).slice(0, 500),
                             from, to, fromTier: assignment.tier, toTier: next.tier, poolBefore,
@@ -3467,14 +3651,15 @@ export async function runDaemon(repoRoot, opts = {}) {
                     // on artificial scarcity; MAX_ATTEMPTS and the review round cap are the real bounds. Only a
                     // fleet whose every channel is DEMOTED (verified dead: auth/setup/provider outage) has nothing
                     // left to spawn — that case alone still parks.
+                    // OBS-1187: shared with the demotion seam's pooled recycle so neither relaunches an escalated adapter.
+                    const recycleExcluded = () => [channelKey(assignment), ...channels.filter((c) => escalatedAdapters.has(c.adapter)).map(channelKey)];
                     const failoverOrRecycle = (site) => {
                         const next = failover(site);
                         if (next)
                             return next;
                         // OBS-1169: a vendor escalated away from stays out of the recycle pool until its release —
                         // recycling any of its channels, launched or not, would launch the broken bootstrap again.
-                        const recycleExcluded = [channelKey(assignment), ...channels.filter((c) => escalatedAdapters.has(c.adapter)).map(channelKey)];
-                        const recycled = nextChannel(assignment, t, cfg, channels, recycleExcluded, profile, demotedChannels)
+                        const recycled = nextChannel(assignment, t, cfg, channels, recycleExcluded(), profile, demotedChannels)
                             ?? (demotedChannels.has(channelKey(assignment)) ? null : assignment);
                         if (recycled)
                             journal.append("channel-recycle", t.id, { site, channel: channelKey(recycled) });
@@ -3493,7 +3678,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                         // D-07: consult panes self-clean when the verdict is read (keepLlm) — only "forever" keeps them.
                         // v1.54 T1: channels = this run's doctor-filtered live list — consult.prefer seat liveness
                         // is judged against it, never rebuilt from config (installed-but-unauthed seats would stall).
-                        { keep: keepLlm, onSlot: keepLlm ? (s) => keptSlots.push(s) : undefined, onInvocation: (inv) => invocations.push(inv), runId, channels: pools.consult });
+                        { keep: keepLlm, onSlot: keepLlm ? (s) => keptSlots.push(s) : undefined, onInvocation: (inv) => invocations.push(inv), runId, channels: pools.consult, health });
                         // A lone answering seat is already the row's adapter/model/vendor/effort; the list is journaled
                         // only when a seat failed, so the row never repeats itself and a failed seat never vanishes.
                         return invocations.some((inv) => inv.outcome === "failed") ? { ...v, invocations } : v;
@@ -3539,7 +3724,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                                     tried.push(k);
                                 return true;
                             }
-                            await park(t, "consult said reroute but every channel is demoted (verified dead) — nothing left to spawn", "reroute-exhausted", assignment, attempts, startMs, gateFails, consults, tokens, metered, retryMode);
+                            await park(t, `consult said reroute but every channel is demoted (verified dead) — nothing left to spawn${poolExhaustion(t, cfg)}`, "reroute-exhausted", assignment, attempts, startMs, gateFails, consults, tokens, metered, retryMode);
                             return false;
                         }
                         await park(t, `consult verdict: ${v.action} — ${v.notes}`, trigger, assignment, attempts, startMs, gateFails, consults, tokens, metered, retryMode); // decompose|human
@@ -3620,20 +3805,19 @@ export async function runDaemon(repoRoot, opts = {}) {
                         // died. A pane the driver cannot read is DECLINED, never classified: an unreadable pane proves
                         // neither a matching trailer nor the absence of a foreign one, so it is neither harvested nor
                         // gated as trailerless evidence. A driver whose terminals outlive it (Orca) adopts the owned
-                        // terminal: bound only on ownership evidence — the journaled owned title AND the task checkout —
-                        // never a title alone. The journaled slot.id is NEVER trusted across daemon instances: a fresh
-                        // driver numbers its slots from one again, so an interrupted task's old id can name whatever
-                        // terminal THIS instance bound under that id (another task's adoption or allocation).
+                        // terminal (Orca) or pane (Herdr, OBS-1203): bound only on ownership evidence — the journaled owned
+                        // name AND the task checkout — never a name alone. The journaled slot.id is NEVER trusted across
+                        // daemon instances: a fresh driver numbers its slots from one again, so an interrupted task's old
+                        // id can name whatever terminal THIS instance bound under that id (another task's adoption or allocation).
                         // A driver with no read-only adoption is declined BEFORE anything is bound: slot() allocates —
                         // herdr's reclaims the same-named pane (closing the evidence unread) and holds a dispatch lease
                         // only run() releases — so it is never an adoption fallback; that driver keeps ordinary recovery.
-                        const adopt = driver.adopt;
-                        if (!adopt)
+                        if (!driver.adopt)
                             return decline("no read-only adoption on this driver");
                         let ownedSlot;
                         let pane;
                         try {
-                            ownedSlot = await adopt.call(driver, slot);
+                            ownedSlot = await driver.adopt(slot);
                             pane = await driver.read(ownedSlot, PANE_READ_ROWS);
                         }
                         catch (error) {
@@ -3661,6 +3845,47 @@ export async function runDaemon(repoRoot, opts = {}) {
                         return found.assignment;
                     };
                     const harvestAuthor = rs && !satisfiedGate && !replayedGates && !recheck ? await harvestInterruptedAttempt() : undefined;
+                    // OBS-1202: a recheck bound to a stall park that recorded a reapFailure re-verifies the parked
+                    // attempt's owned census through the same reapWorker a live harvest uses, BEFORE any gate reads
+                    // its tree. Only a complete, empty census releases the harvested commits to the battery; an
+                    // unreadable (null) or surviving census keeps the task parked on the same kind — a stall carrying
+                    // the fresh reapFailure, bound by its own new token — and no worker is funded either way. Plain
+                    // approve on that park keeps its dispatch semantics.
+                    if (recheck) {
+                        const approvalRow = effectiveEvents(journal.read()).reverse()
+                            .find((e) => e.taskId === t.id && e.event === "task-approved" && e.data.release === RECHECK_RELEASE);
+                        if (typeof approvalRow?.data.reapFailure === "string") {
+                            const launch = [...journal.read()].reverse().find((e) => e.taskId === t.id && e.event === "worker-launch");
+                            const owned = launch?.data.slot;
+                            let uncertain;
+                            if (typeof launch?.data.nonce !== "string" || typeof launch.data.dispatchScript !== "string"
+                                || typeof owned?.name !== "string" || typeof owned.cwd !== "string") {
+                                uncertain = "the parked attempt recorded no launch ownership evidence";
+                            }
+                            else {
+                                const key = { ...owned };
+                                workerOwners.set(key, { taskId: t.id, attempt: Number(launch.data.attempt ?? 0),
+                                    groupFile: `${launch.data.dispatchScript}.${launch.data.nonce}.pgid`, marker: launch.data.dispatchScript,
+                                    identities: new Map(), descendants: new Map() });
+                                try {
+                                    await reapWorker(key);
+                                    // reapWorker returns normally for a null census when no ownership file was readable (a
+                                    // missing .pgid/.session with no descendants); recovery needs a RECORDED empty census.
+                                    const census = reapReports.get(key)?.survivors;
+                                    if (!Array.isArray(census) || census.length > 0) {
+                                        uncertain = `worker group ${reapReports.get(key)?.processGroup ?? "unknown"} census unrecorded`;
+                                    }
+                                }
+                                catch (error) {
+                                    uncertain = error instanceof Error ? error.message : String(error);
+                                }
+                            }
+                            if (uncertain !== undefined) {
+                                await park(t, `worker could not be reaped before recheck: ${uncertain}`, "stall", null, rs?.attempts ?? 0, startMs, gateFails, consults, tokens, metered, retryMode, { reapFailure: uncertain });
+                                return;
+                            }
+                        }
+                    }
                     resumeGateReplay: if (satisfiedGate || replayedGates || recheck || harvestAuthor) {
                         const taskBase = await integrationHead(intWt);
                         taskBases.set(t.id, taskBase);
@@ -3887,14 +4112,27 @@ export async function runDaemon(repoRoot, opts = {}) {
                             }
                             await trackedDriver.project?.(t.id, "in-review");
                             await waitForBaseline(t.id);
+                            // OBS-1106 residual: the live loop's timeout refresh, decided again from the journal, so a crash
+                            // cannot launder the copy it forbids. A harvested retry that landed nothing owes it exactly as
+                            // the live loop would have; a crash inside a forced battery before its test row still owes it.
+                            const lastRound = (() => {
+                                const own = journal.read().filter((e) => e.taskId === t.id);
+                                return own.slice(own.map((e) => e.event === "phase-start" && e.data.phase === "gates").lastIndexOf(true) + 1);
+                            })();
+                            const freshTimeout = !recheck && !satisfiedGate && ((harvestAuthor !== undefined && owesTimeoutRefresh(t, lastRound, gateSubject.commit))
+                                || (lastRound.some((e) => e.event === "gate-fresh-forced" && e.data.commit === gateSubject.commit)
+                                    && !lastRound.some((e) => e.event === "gate-result" && e.data.gate === "test")));
                             journal.phaseStart(t.id, "gates");
+                            if (freshTimeout) {
+                                journal.append("gate-fresh-forced", t.id, { gate: "test", attempt: gateSubject.attempt, commit: gateSubject.commit, reason: "no-commit-timeout-red", resumed: true });
+                            }
                             const { results } = await withCommandContext(t.id, async () => runReviewRecovery(resumedTask, {
                                 carriedAuthors: await subjectAuthors(journal.read(), t.id, wt, taskBase),
                                 producer: producerNow(),
                                 carriedFindings: outstandingReviewFindings(journal.read(), t.id),
                                 operatorContext,
                                 worktree: wt, baseRef: taskBase, result: priorResult, author: gateAuthor,
-                                commands, baseline, channels: pools.review, judgeChannels: pools.judge, adapters, cfg, artifactDir: journal.dir,
+                                commands, baseline, channels: pools.review, judgeChannels: pools.judge, health, adapters, cfg, artifactDir: journal.dir,
                                 collateral: collateral.get(t.id) ?? [],
                                 // a recheck re-verifies a human's release: it never selects tests down, it runs the suite.
                                 via: cfg.visibility.llm === "pane"
@@ -3912,7 +4150,10 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 // flaking seat is re-asked on every task.
                                 reviewNoVerdicts,
                                 recheck, // OBS-1055: a recheck discards cached reds — the battery re-measures what the operator questioned
+                                cachedRedBypass: freshTimeout ? "timeout-fresh" : undefined,
                                 onGate: async (e) => {
+                                    if (!liveEngagement())
+                                        return;
                                     if (e.phase === "start") {
                                         notePhaseStart(e);
                                         journal.phaseStart(t.id, phaseForGate(e.gate), { gate: e.gate, index: e.index, total: e.total, ...(e.parentAt === undefined ? {} : { parallel: true }) });
@@ -3928,10 +4169,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                                     inParallelOrder(g.gate, () => {
                                         journalGateResult(g);
                                         noteReviewRetry(g);
-                                        if (g.gate === "review" && !g.pass && /unparseable/.test(g.details)
-                                            && typeof g.meta?.reviewer === "string") {
+                                        if (isGarbageReview(g))
                                             badReviewers.push(g.meta.reviewer);
-                                        }
                                     });
                                 },
                             }, false));
@@ -3974,6 +4213,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 // erase an EARLIER chargeable attempt and attribute the park to the assignment the rewind
                                 // restored. Ask the journal which state this is, and take the classified dispatch's own
                                 // assignment back with it.
+                                if (await adjudicateInfraShapedRed(t, results, wt, gateSubject.commit, rs?.attempts ?? 0, false, (reason) => park(t, reason, "infra", gateAuthor, rs?.attempts ?? 0, startMs, gateFails, consults, tokens, metered, retryMode)))
+                                    return;
                                 const classified = journal.classifiedDispatch(t.id);
                                 if (await dispositionScopeRed(t, results, classified?.assignment ?? gateAuthor, classified ? (rs?.attempts ?? 0) : Math.max(0, (rs?.attempts ?? 0) - 1), startMs, gateFails, consults, tokens, metered, retryMode, await treeOrDiffPaths(taskBase, wt)))
                                     return;
@@ -4103,13 +4344,21 @@ export async function runDaemon(repoRoot, opts = {}) {
                         }
                         // OBS-57: a demoted channel must not be re-dispatched on consult retry or provider requeue.
                         if (demotedChannels.has(channelKey(assignment))) {
-                            const next = nextChannel(assignment, t, cfg, channels, tried, profile, demotedChannels);
+                            // OBS-1187: a pooled task recycles a live pool member (OBS-202) before it may park naming the pool;
+                            // it never keeps a demoted seat and never leaves the pool.
+                            const pooled = taskPool(t, cfg) !== undefined;
+                            const next = nextChannel(assignment, t, cfg, channels, tried, profile, demotedChannels)
+                                ?? (pooled ? nextChannel(assignment, t, cfg, channels, recycleExcluded(), profile, demotedChannels) : null);
                             if (next) {
                                 assignment = next;
                                 const k = channelKey(next);
                                 if (!tried.includes(k))
                                     tried.push(k);
                             }
+                            else if (pooled) {
+                                await park(t, `${channelKey(assignment)} is demoted (verified dead) and no live pool member remains${poolExhaustion(t, cfg)}`, "reroute-exhausted", assignment, attempt, startMs, gateFails, consults, tokens, metered, retryMode);
+                                return;
+                            }
                         }
                         // v1.29: consume the prior gate-failed session once. Same channel + known under-threshold context
                         // + adapter capability resumes; every other path is today's fresh dispatch.
@@ -4134,7 +4383,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 gate: banned, from: channelKey(assignment), to: next ? channelKey(next) : null,
                             });
                             if (!next) {
-                                await park(t, gateFailApprovalReason(t.id, `identical ${banned} failure twice this engagement — an identical retry is banned and no untried channel is left`), "gate-fail", assignment, attempt, startMs, gateFails, consults, tokens, metered, retryMode);
+                                await park(t, gateFailApprovalReason(t.id, `identical ${banned} failure twice this engagement — an identical retry is banned and no untried channel is left${poolExhaustion(t, cfg)}`), "gate-fail", assignment, attempt, startMs, gateFails, consults, tokens, metered, retryMode);
                                 return;
                             }
                             assignment = next;
@@ -4348,10 +4597,18 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 ? `${new TextDecoder().decode(rawBytes.subarray(0, cap))}\n… diff truncated at gates.diffCap (${cap} bytes)`
                                     + `\nPARTIAL DIFF: the diff above is incomplete; files cut by the cap: ${droppedFiles.join(", ") || "(unknown)"}`
                                 : raw.stdout;
-                            const brief = repairBrief(repairFindings, diff, taskBase);
+                            const repeated = reraisedReviewChains(journaledSoFar, t.id).filter((chain) => chain.reraises > 1);
+                            const brief = repairBrief(repairFindings, diff, taskBase, repeated);
                             // anything the live brief holds beyond the journaled findings (a consult's guidance) is kept:
                             // a repair adds the diff and the fix-only contract, it never subtracts what was already known.
-                            feedback = feedback && !brief.includes(feedback) ? `${brief}\n\n${feedback}` : brief;
+                            // OBS-1195: a resumed upheld repair re-derives the same review bytes on both carries; when the
+                            // repair already quotes them verbatim they are said once, and the operator's uphold line stays.
+                            // Both carries had their deferrals cut above, so the quote is compared after the same cut.
+                            const upheldQuote = upheldFeedback ? withoutDeferrals(`review: ${upheldFeedback}`) : undefined;
+                            const rest = upheldQuote && brief.includes(upheldQuote)
+                                ? feedback.replace(upheldQuote, "").replace(/\n{3,}/g, "\n\n").trim()
+                                : feedback;
+                            feedback = rest && !brief.includes(rest) ? `${brief}\n\n${rest}` : brief;
                             journal.append("repair-dispatch", t.id, { workerDispatchOrdinal, diffBytes: Buffer.byteLength(diff, "utf8"), capped, ...(capped ? { droppedFiles } : {}) });
                         }
                         if (feedback || priorNamed.length > 0) {
@@ -5675,7 +5932,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 attempt--; // the dead bootstrap bought no worker turn
                                 continue;
                             }
-                            await park(t, `${from} failed at bootstrap after ${retries} same-channel retry and no eligible channel remains`, "infra", assignment, attempt, startMs, gateFails, consults, tokens, metered, retryMode, { cause: "bootstrap", channel: from, retries });
+                            await park(t, `${from} failed at bootstrap after ${retries} same-channel retry and no eligible channel remains${poolExhaustion(t, cfg)}`, "infra", assignment, attempt, startMs, gateFails, consults, tokens, metered, retryMode, { cause: "bootstrap", channel: from, retries });
                             return;
                         }
                         const preHarvestResult = result;
@@ -5787,7 +6044,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 tried.push(channelKey(next));
                                 continue;
                             }
-                            await park(t, `capacity exhausted on ${from} after ${requeues} requeues and no eligible channel at floor`, "quota", assignment, attempt + 1, startMs, gateFails, consults, tokens, metered, retryMode, { cause: "capacity", channel: from, requeues });
+                            await park(t, `capacity exhausted on ${from} after ${requeues} requeues and no eligible channel at floor${poolExhaustion(t, cfg)}`, "quota", assignment, attempt + 1, startMs, gateFails, consults, tokens, metered, retryMode, { cause: "capacity", channel: from, requeues });
                             return;
                         }
                         // quota exhaustion → failover within floor; does NOT consume the ladder (spec §4)
@@ -5850,7 +6107,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 tried.push(channelKey(next));
                                 continue;
                             }
-                            await park(t, "quota exhausted on every eligible channel", "quota", assignment, attempt + 1, startMs, gateFails, consults, tokens, metered, retryMode);
+                            await park(t, `quota exhausted on every eligible channel${poolExhaustion(t, cfg)}`, "quota", assignment, attempt + 1, startMs, gateFails, consults, tokens, metered, retryMode);
                             return;
                         }
                         // v1.65 T1: typed dead-channel failure — the parse boundary classified this no-trailer result
@@ -5898,7 +6155,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 continue;
                             }
                             const chargedAttempts = adapterStartupFailure || startupFailure ? attempt : attempt + 1;
-                            await park(t, `dead channel (${dead}) and no eligible channel remains`, "reroute-exhausted", assignment, chargedAttempts, startMs, gateFails, consults, tokens, metered, retryMode);
+                            await park(t, `dead channel (${dead}) and no eligible channel remains${poolExhaustion(t, cfg)}`, "reroute-exhausted", assignment, chargedAttempts, startMs, gateFails, consults, tokens, metered, retryMode);
                             return;
                         }
                         if (!finished) {
@@ -5943,6 +6200,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                             }
                         }
                         const onGate = async (e) => {
+                            if (!liveEngagement())
+                                return;
                             if (e.phase === "start") {
                                 notePhaseStart(e);
                                 journal.phaseStart(t.id, phaseForGate(e.gate), { gate: e.gate, index: e.index, total: e.total, ...(e.parentAt === undefined ? {} : { parallel: true }) });
@@ -5976,9 +6235,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 journalGateResult(g);
                                 noteReviewRetry(g);
                                 // v1.1 failover: never re-ask a reviewer channel that produced garbage for this task
-                                if (g.gate === "review" && !g.pass && /unparseable/.test(g.details) && typeof g.meta?.reviewer === "string") {
+                                if (isGarbageReview(g))
                                     badReviewers.push(g.meta.reviewer);
-                                }
                             });
                         };
                         let results = [];
@@ -5994,12 +6252,16 @@ export async function runDaemon(repoRoot, opts = {}) {
                             // sees the replay, without buying another command or reviewer invocation.
                             const taskEvents = journal.read().filter((e) => e.taskId === t.id);
                             const previousRound = taskEvents.map((e) => e.event === "phase-start" && e.data.phase === "gates").lastIndexOf(true);
-                            const previousRows = retryMode === "repair"
-                                ? taskEvents.slice(previousRound + 1).filter((e) => e.event === "gate-result"
-                                    && e.data.attempt === attempt - 1)
-                                : [];
+                            const lastRoundRows = taskEvents.slice(previousRound + 1).filter((e) => e.event === "gate-result"
+                                && e.data.attempt === attempt - 1);
+                            const previousRows = retryMode === "repair" ? lastRoundRows : [];
+                            // OBS-1106 residual: see owesTimeoutRefresh — the replay below and the verdict cache are both refused.
+                            const freshTimeout = !explicitlyFundedAttempt && owesTimeoutRefresh(t, lastRoundRows, gateSubject.commit);
                             journal.phaseStart(t.id, "gates");
-                            const replay = !explicitlyFundedAttempt && previousRows.length > 0
+                            if (freshTimeout) {
+                                journal.append("gate-fresh-forced", t.id, { gate: "test", attempt, priorAttempt: attempt - 1, commit: gateSubject.commit, reason: "no-commit-timeout-red" });
+                            }
+                            const replay = !explicitlyFundedAttempt && !freshTimeout && previousRows.length > 0
                                 && previousRows.every((e) => e.data.commit === gateSubject.commit)
                                 && previousRows.some((e) => e.data.pass === false && e.data.skipped !== true);
                             if (replay) {
@@ -6029,10 +6291,10 @@ export async function runDaemon(repoRoot, opts = {}) {
                                     carriedFindings: outstandingFindings,
                                     operatorContext,
                                     worktree: wt, baseRef: taskBase, result, author: assignment,
-                                    commands, baseline, channels: pools.review, judgeChannels: pools.judge, adapters, cfg, artifactDir: journal.dir,
+                                    commands, baseline, channels: pools.review, judgeChannels: pools.judge, health, adapters, cfg, artifactDir: journal.dir,
                                     collateral: collateral.get(t.id) ?? [],
                                     selectTests: !testGateFailed,
-                                    cachedRedBypass: explicitlyFundedAttempt ? "operator-rerun" : undefined,
+                                    cachedRedBypass: explicitlyFundedAttempt ? "operator-rerun" : freshTimeout ? "timeout-fresh" : undefined,
                                     via: cfg.visibility.llm === "pane"
                                         ? {
                                             driver: trackedDriver,
@@ -6139,6 +6401,8 @@ export async function runDaemon(repoRoot, opts = {}) {
                             await park(t, `${capTrip.gate}: ${capTrip.details}`, "diff-cap", assignment, attempt + 1, startMs, gateFails, consults, tokens, metered, retryMode);
                             return;
                         }
+                        if (await adjudicateInfraShapedRed(t, results, wt, gateSubject.commit, attempt, gateSubject.replayedFromAttempt !== undefined, (reason) => park(t, reason, "infra", assignment, attempt + 1, startMs, gateFails, consults, tokens, metered, retryMode)))
+                            return;
                         // OBS-547: who pays for this red is decided by the run's collateral prediction (see
                         // dispositionScopeRed) — an authoring defect parks unchargeable before any accounting below.
                         if (await dispositionScopeRed(t, results, assignment, attempt, startMs, gateFails, consults, tokens, metered, retryMode, await treeOrDiffPaths(taskBase, wt)))
@@ -6371,6 +6635,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                         for (const t of ready) {
                             const p = executeWithPolicy(t)
                                 .catch(async (err) => {
+                                engagements.delete(t.id);
                                 const cleanupErrors = [];
                                 for (const [slot, owner] of workerOwners) {
                                     if (owner.taskId !== t.id || !liveSlots.has(slot))
@@ -6429,7 +6694,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                                 journal.telemetry({ taskId: t.id, shape: t.shape, adapter: "-", model: "-", channel: "-", attempts: 0, outcome: "failed", durationMs: 0 });
                                 await reconcile({ spareLiveLlm: true }); // task-failed is a terminal event
                             })
-                                .finally(() => inflight.delete(t.id));
+                                .finally(() => { inflight.delete(t.id); engagements.delete(t.id); });
                             inflight.set(t.id, p);
                         }
                         if (inflight.size === 0) {
@@ -6529,7 +6794,7 @@ export async function runDaemon(repoRoot, opts = {}) {
                         poll = setTimeout(pollApprovals, APPROVAL_POLL_MS);
                         let tipFailed;
                         try {
-                            const verification = withCommandContext(undefined, () => verifyIntegrationTipCached(intWt, commands, journal, { lastMergedTask, baseline, signal: controller.signal }), controller.signal);
+                            const verification = withCommandContext(undefined, () => verifyIntegrationTipCached(intWt, commands, journal, { lastMergedTask, baseline, signal: controller.signal, evidence: { quotaBytes: cfg.gates?.evidenceQuotaBytes } }), controller.signal);
                             activeTipVerify = { controller, settled: verification.then(() => { }, () => { }) };
                             tipFailed = await verification;
                         }