audit-tools 0.52.2 → 0.52.4

package/dist/audit/cli/dispatch/hostHandoff.js

@@ -1,3 +1,4 @@
+import { recordHostRootLogBoundary } from "../../../shared/observability/rootLogObservations.js";
 // sites-pinned: tests/audit/host-handoff.test.ts
 // (the steward-lane verification contract: the lane-aware prompt, the one
 // optional envelope key, the refusal of that key on a lane whose contract never
@@ -5,7 +6,7 @@
 // task-bindings version bump whose refusal names the remedy)
 import { mkdir } from "node:fs/promises";
 import { join } from "node:path";
-import { createMemoizedSourceReader, appendSubmissionEvent, SUBMISSION_LEDGER_EVENT_CONTRACT_VERSION, bindingIdentity, compareCodeUnits, contentSha256, firstDuplicateIdentity, hasExactKeys, hostHandoffResultPath, identityFailureDiagnostic, isFileMissingError, isRecord, isSha256, LaneDemandSchema, parseAllWorkloadItems, parseWorkloadEnvelope, bindWorkerPrompt, deriveResultId, workerPromptBindingHolds, enrichMissingSubmissionIssues, readTrailingSubmissionRefusals, readJsonFile, repoRelativePath, requireNonEmptyString, resolveContainedPath, resolveHostHandoffPaths, resultMapIdentity, scanBoundSubmission, siblingLockPath, sameStrings, stableStringify, verifyFindingGrounding, withFileLock, writeBlockedStepContract, writeJsonFile, } from "audit-tools/shared";
+import { createMemoizedSourceReader, appendSubmissionEvent, SUBMISSION_LEDGER_EVENT_CONTRACT_VERSION, bindingIdentity, compareCodeUnits, contentSha256, firstDuplicateIdentity, hasExactKeys, hostHandoffResultPath, identityFailureDiagnostic, isFileMissingError, isJsonParseError, isRecord, isSha256, LaneDemandSchema, parseAllWorkloadItems, parseWorkloadEnvelope, bindWorkerPrompt, deriveResultId, workerPromptBindingHolds, enrichMissingSubmissionIssues, readTrailingSubmissionRefusals, readJsonFile, repoRelativePath, requireNonEmptyString, resolveContainedPath, resolveHostHandoffPaths, resultMapIdentity, scanBoundSubmission, siblingLockPath, sameStrings, stableStringify, verifyFindingGrounding, withFileLock, writeBlockedStepContract, writeJsonFile, } from "audit-tools/shared";
 import { findingContractPromptLines } from "../../contracts/findingContractPrompt.js";
 import { WORKER_REFUSED_FINDING_VERDICTS, WorkerFindingSchema, } from "../../contracts/workerSchemas.js";
 import { AuditResultSchema, AuditTaskSchema, } from "../../types.js";
@@ -142,7 +143,7 @@ function resolveBoundaryPaths(params) {
     };
 }
 /**
- * The accepted-results read-modify-write, SERIALIZED.
+ * The binding publication, ingestion reads, and accepted-results write, SERIALIZED.
  *
  * `host-accepted-results.json` and its ledger are one logical record written as
  * two files, and both prepare and ingest used to load a snapshot, work from it,
@@ -151,7 +152,9 @@ function resolveBoundaryPaths(params) {
  * predates the earlier writer's additions, and every duplicate-binding and
  * duplicate-result_id guard downstream derives from that stale snapshot.
  *
- * So the read, the merge and both writes happen inside ONE acquisition of the
+ * The workload/bindings/map publish and ingest read set also share this hold,
+ * so a concurrent prepare cannot replace a binding during acceptance.
+ * The read, the merge and both writes happen inside ONE acquisition of the
  * shared lock substrate. No backoff, retry or stale-lock logic lives here — all
  * of it is `withFileLock`'s, and the caller's RunLogger is threaded straight
  * through so the primitive's heartbeat and stale-lock-reclaim events land in the
@@ -342,8 +345,8 @@ function buildPrompt(task, resultPath, workloadPath) {
             // Required is the honest reading: a steward returns `findings: []`, so a
             // steward with no `verification` carries no answer at all and is
             // indistinguishable from a lane that failed.
-            ? "Result contract: audit-host-result/v1alpha1 with exactly result_id, run_id, work_item_id, prompt_sha256, file_coverage, findings, reviewed_clean, and verification in addition to contract_version. On this lane verification is REQUIRED."
-            : "Result contract: audit-host-result/v1alpha1 with exactly result_id, run_id, work_item_id, prompt_sha256, file_coverage, and findings in addition to contract_version, plus reviewed_clean when findings is empty.",
+            ? "Result contract: audit-host-result/v1alpha1 with exactly result_id, run_id, work_item_id, prompt_sha256, file_coverage, findings, and verification in addition to contract_version. On this lane verification is REQUIRED."
+            : "Result contract: audit-host-result/v1alpha1 with exactly result_id, run_id, work_item_id, prompt_sha256, file_coverage, and findings in addition to contract_version.",
         "Each file_coverage entry must contain exactly path, reviewed_lines, and total_lines.",
         // The finding contract is CARRIED, not referenced: it is rendered from the
         // very schema ingestion enforces, so a host never has to remember or fetch it.
@@ -640,20 +643,6 @@ export async function prepareAuditHostHandoff(params) {
         }),
     };
     await mkdir(paths.resultDir, { recursive: true });
-    await writeJsonFile(paths.taskBindingsPath, taskBindings);
-    // Say out loud whether this wave has drained. The caller hands in the run's
-    // still-OWED partition, so an EMPTY one is the wave's own statement that every
-    // work item it published has been accepted — and that is the only evidence
-    // there is: from the review-run resolution, a fully-accepted wave and a wave
-    // with one lane still out look identical (both are "fewer pending than
-    // published"). The next wave's identity is derived from this, so it is written
-    // where the run itself lives and never inferred. A comment stating this does
-    // NOT survive; the marker is what does.
-    await writeJsonFile(reviewWaveClosedPath(params.artifactsDir, params.runId), {
-        contract_version: REVIEW_WAVE_CLOSED_CONTRACT_VERSION,
-        run_id: params.runId,
-        closed: allWorkItems.length === 0,
-    });
     // What is PUBLISHED is every task this caller handed in. This boundary does
     // not suppress work items the accepted ledger names, and deliberately so —
     // the ledger records what ARRIVED, it never decides what is still OWED.
@@ -668,6 +657,22 @@ export async function prepareAuditHostHandoff(params) {
     // acquisition: a prepare that snapshotted before a concurrent ingest's
     // additions must not replace them with its stale copy.
     return withAcceptedResultsLock(paths, params.logger, async (accepted) => {
+        if (allWorkItems.length > 0)
+            await recordHostRootLogBoundary({ ...paths, phase: "prepare", logger: params.logger });
+        await writeJsonFile(paths.taskBindingsPath, taskBindings);
+        // Say out loud whether this wave has drained. The caller hands in the run's
+        // still-OWED partition, so an EMPTY one is the wave's own statement that every
+        // work item it published has been accepted — and that is the only evidence
+        // there is: from the review-run resolution, a fully-accepted wave and a wave
+        // with one lane still out look identical (both are "fewer pending than
+        // published"). The next wave's identity is derived from this, so it is written
+        // where the run itself lives and never inferred. A comment stating this does
+        // NOT survive; the marker is what does.
+        await writeJsonFile(reviewWaveClosedPath(params.artifactsDir, params.runId), {
+            contract_version: REVIEW_WAVE_CLOSED_CONTRACT_VERSION,
+            run_id: params.runId,
+            closed: allWorkItems.length === 0,
+        });
         const workload = {
             contract_version: WORKLOAD_CONTRACT_VERSION,
             run_id: params.runId,
@@ -863,7 +868,7 @@ function parseResultMap(value, runId) {
             typeof entry.work_item_id === "string" &&
             isSha256(entry.prompt_sha256) &&
             typeof entry.result_path === "string")) {
-        throw new Error("Invalid audit host result map");
+        throw bindingFailure("workload_binding", "Invalid audit host result map");
     }
     return value;
 }
@@ -911,7 +916,7 @@ function parseTaskBindings(value, runId) {
         !hasExactKeys(value, ["contract_version", "entries", "run_id"]) ||
         value.run_id !== runId ||
         !Array.isArray(value.entries)) {
-        throw new Error("Invalid audit host task bindings");
+        throw bindingFailure("workload_binding", "Invalid audit host task bindings");
     }
     // The VERSION is checked on its OWN, before the shape walk, so a binding set
     // written under an older version is refused for THAT reason and carries the
@@ -927,7 +932,7 @@ function parseTaskBindings(value, runId) {
     for (const rawEntry of value.entries) {
         const entry = parseTaskBinding(rawEntry);
         if (entry === null || bindings.has(entry.work_item_id)) {
-            throw new Error("Invalid or duplicate audit host task binding");
+            throw bindingFailure("workload_binding", "Invalid or duplicate audit host task binding");
         }
         bindings.set(entry.work_item_id, entry);
     }
@@ -966,14 +971,19 @@ function validateHandoffBinding(paths, workload, resultMap, taskBindings) {
     }
     return items;
 }
-/**
- * A persisted-binding failure is a THROW, not an issue: nothing about a host
- * result can be judged until the tool's own workload, result map and task
- * bindings re-derive. The check id is what the throw cites, so the failure is
- * still attributable to a registered ingestion check.
- */
+/** Recognized tool-owned binding corruption; the owning emitter can re-prepare it. */
+class AuditHostBindingError extends Error {
+    check;
+    code = "workload_stale";
+    constructor(check, message) {
+        super(`${message} [${check}]; re-prepare the handoff to restore its canonical binding set. ` +
+            "No submission was accepted against these invalid bindings.");
+        this.check = check;
+        this.name = "AuditHostBindingError";
+    }
+}
 function bindingFailure(check, message) {
-    return new Error(`${message} [${check}]`);
+    return new AuditHostBindingError(check, message);
 }
 /** `check` names the registered ingestion check that failed; `detail` is its prose. */
 function refuse(check, detail) {
@@ -1225,228 +1235,220 @@ function auditScanMessages(workItemId) {
 }
 export async function ingestAuditHostResults(params) {
     const paths = resolveBoundaryPaths(params);
-    const accepted = await loadAcceptedResults(paths.acceptedLedgerPath, params.runId);
-    // A STALE persisted document is refused as a CLASSIFIED ISSUE, never as a
-    // throw. These are the refusals with a named repair — re-prepare, which this
-    // fold performs on its way to the next emission — so they must reach the host
-    // as rendered diagnostics rather than as unclassified stacks out of the fold.
-    // Nothing is accepted: a document this build did not mint cannot be
-    // re-derived, so no submission has a binding to be judged against.
-    //
-    // The BINDING SET is in this class for one more reason than the workload is:
-    // on the fold, a throw here lands BEFORE the one path that re-prepares, so it
-    // wedges the run rather than surfacing anything (see
-    // {@link StaleAuditHostTaskBindingsError}). A result the host already wrote
-    // under the old bindings is therefore REFUSED as stale — it is never accepted
-    // against the new contract, because the new contract's binding is what the
-    // re-prepare mints, and the item is re-published under it.
-    const stale = (error) => ({
-        accepted_count: 0,
-        accepted_results: accepted.entries.map((entry) => entry.audit_result),
-        accepted_results_path: paths.acceptedResultsPath,
-        completed_work_item_ids: [
-            ...new Set(accepted.entries.map((entry) => entry.work_item_id)),
-        ].sort(compareCodeUnits),
-        issues: [error],
-        raw_issues: [error],
-        validation_warnings: [],
-    });
-    let workload;
-    try {
-        workload = parseWorkload(await readJsonFile(paths.workloadPath), params.runId);
-    }
-    catch (error) {
-        if (!(error instanceof StaleAuditHostWorkloadError))
-            throw error;
-        return stale(error);
-    }
-    const resultMap = parseResultMap(await readJsonFile(paths.resultMapPath), params.runId);
-    let taskBindings;
-    try {
-        taskBindings = parseTaskBindings(await readJsonFile(paths.taskBindingsPath), params.runId);
-    }
-    catch (error) {
-        if (!(error instanceof StaleAuditHostTaskBindingsError))
-            throw error;
-        return stale(error);
-    }
-    const items = validateHandoffBinding(paths, workload, resultMap, taskBindings);
-    const acceptedBindings = new Set(accepted.entries.map(bindingIdentity));
-    const resultIds = new Set(accepted.entries.map((entry) => entry.result_id));
-    const additions = [];
-    const issues = [];
-    const validation_warnings = [];
-    // One memoized reader for the whole ingest: N findings citing one file read it once.
-    const readSource = createMemoizedSourceReader();
-    const activeTaskIds = new Set(params.auditTasks.map((task) => task.task_id));
-    for (const entry of resultMap.entries) {
-        if (acceptedBindings.has(bindingIdentity(entry)))
-            continue;
-        const item = items.get(entry.work_item_id);
-        const binding = taskBindings.get(entry.work_item_id);
-        if (item === undefined || binding === undefined)
-            continue;
-        const outcome = await scanBoundSubmission({
-            root: paths.root,
-            artifactsDir: paths.artifactsDir,
-            workItemId: entry.work_item_id,
-            resultPath: entry.result_path,
-            parse: (value) => {
-                const parsed = parseHostResult(value, params.runId, item, binding);
-                return parsed.ok ? { ok: true, parsed: parsed.result } : parsed;
-            },
-            resultId: (result) => result.result_id,
-            // CHECK only. The id is consumed further down, after conversion,
-            // validation and grounding — a result refused there must stay re-submittable.
-            seen: (resultId) => resultIds.has(resultId),
-            messages: auditScanMessages(entry.work_item_id),
+    return withAcceptedResultsLock(paths, params.logger, async (accepted) => {
+        // A STALE persisted document is refused as a CLASSIFIED ISSUE, never as a
+        // throw. These are the refusals with a named repair — re-prepare, which this
+        // fold performs on its way to the next emission — so they must reach the host
+        // as rendered diagnostics rather than as unclassified stacks out of the fold.
+        // Nothing is accepted: a document this build did not mint cannot be
+        // re-derived, so no submission has a binding to be judged against.
+        //
+        // The BINDING SET is in this class for one more reason than the workload is:
+        // on the fold, a throw here lands BEFORE the one path that re-prepares, so it
+        // wedges the run rather than surfacing anything (see
+        // {@link StaleAuditHostTaskBindingsError}). A result the host already wrote
+        // under the old bindings is therefore REFUSED as stale — it is never accepted
+        // against the new contract, because the new contract's binding is what the
+        // re-prepare mints, and the item is re-published under it.
+        const stale = (error) => ({
+            accepted_count: 0,
+            accepted_results: accepted.entries.map((entry) => entry.audit_result),
+            accepted_results_path: paths.acceptedResultsPath,
+            completed_work_item_ids: [
+                ...new Set(accepted.entries.map((entry) => entry.work_item_id)),
+            ].sort(compareCodeUnits),
+            issues: [error],
+            raw_issues: [error],
+            validation_warnings: [],
         });
-        if (!outcome.ok) {
-            issues.push(outcome.issue);
-            continue;
+        let workload;
+        let resultMap;
+        let taskBindings;
+        let items;
+        try {
+            workload = parseWorkload(await readJsonFile(paths.workloadPath), params.runId);
+            resultMap = parseResultMap(await readJsonFile(paths.resultMapPath), params.runId);
+            taskBindings = parseTaskBindings(await readJsonFile(paths.taskBindingsPath), params.runId);
+            items = validateHandoffBinding(paths, workload, resultMap, taskBindings);
         }
-        const result = outcome.parsed;
-        const converted = toAuditResult(result, binding);
-        if (!converted.ok) {
-            issues.push({
-                code: "submission_contract_invalid",
-                check: "result_schema",
-                message: `work item '${entry.work_item_id}' submitted a result that does not convert to ` +
-                    `an AuditResult: ${converted.detail}`,
-                work_item_id: entry.work_item_id,
-                result_path: entry.result_path,
-            });
-            continue;
-        }
-        // VALIDATE BEFORE ACCEPT. The conversion above proves only the envelope
-        // contract (`FindingSchema` admits an evidence-less finding); these are the
-        // per-result rules the downstream batch gate applies, applied HERE so an
-        // error-severity issue never reaches the accepted pair. A rejected item is
-        // simply never in the ledger, so the corrected file at the same bound path
-        // is re-read on the next fold — acceptance used to be terminal instead, and
-        // a failed batch gate then wedged the run permanently.
-        //
-        // Orphans (task pruned by a re-plan) pass through UNVALIDATED with the same
-        // stderr notice the batch gate uses — never newly rejected: refusing one
-        // here would strand it outside the append-only ledger entirely.
-        if (!activeTaskIds.has(entry.work_item_id)) {
-            process.stderr.write(`audit host-handoff ingest: result for '${entry.work_item_id}' is not in the ` +
-                `active task manifest (orphaned by re-planning); retained in the accepted pair ` +
-                `but skipped at the validation gate\n`);
+        catch (error) {
+            if (error instanceof StaleAuditHostWorkloadError ||
+                error instanceof StaleAuditHostTaskBindingsError ||
+                error instanceof AuditHostBindingError)
+                return stale(error);
+            if (isJsonParseError(error)) {
+                return stale(new AuditHostBindingError("workload_binding", error.message));
+            }
+            // Infrastructure failures and corrupt accepted ledgers remain strict.
+            throw error;
         }
-        else {
-            const validationIssues = validateOneAuditResult(converted.auditResult, [
-                ...params.auditTasks,
-            ], {
-                lineIndex: params.lineIndex,
+        const acceptedBindings = new Set(accepted.entries.map(bindingIdentity));
+        const resultIds = new Set(accepted.entries.map((entry) => entry.result_id));
+        const additions = [];
+        const issues = [];
+        const validation_warnings = [];
+        // One memoized reader for the whole ingest: N findings citing one file read it once.
+        const readSource = createMemoizedSourceReader();
+        const activeTaskIds = new Set(params.auditTasks.map((task) => task.task_id));
+        for (const entry of resultMap.entries) {
+            if (acceptedBindings.has(bindingIdentity(entry)))
+                continue;
+            const item = items.get(entry.work_item_id);
+            const binding = taskBindings.get(entry.work_item_id);
+            if (item === undefined || binding === undefined)
+                continue;
+            const outcome = await scanBoundSubmission({
+                root: paths.root,
+                artifactsDir: paths.artifactsDir,
+                workItemId: entry.work_item_id,
+                resultPath: entry.result_path,
+                parse: (value) => {
+                    const parsed = parseHostResult(value, params.runId, item, binding);
+                    return parsed.ok ? { ok: true, parsed: parsed.result } : parsed;
+                },
+                resultId: (result) => result.result_id,
+                // CHECK only. The id is consumed further down, after conversion,
+                // validation and grounding — a result refused there must stay re-submittable.
+                seen: (resultId) => resultIds.has(resultId),
+                messages: auditScanMessages(entry.work_item_id),
             });
-            const errors = validationIssues.filter((issue) => issue.severity === "error");
-            if (errors.length > 0) {
+            if (!outcome.ok) {
+                issues.push(outcome.issue);
+                continue;
+            }
+            const result = outcome.parsed;
+            const converted = toAuditResult(result, binding);
+            if (!converted.ok) {
                 issues.push({
-                    code: "result_validation_failed",
-                    check: "result_validation",
-                    message: `work item '${entry.work_item_id}' failed audit-results validation ` +
-                        `(${errors.length} error(s)); fix the result file at its bound path and call next-step again: ` +
-                        formatAuditResultIssues(errors),
+                    code: "submission_contract_invalid",
+                    check: "result_schema",
+                    message: `work item '${entry.work_item_id}' submitted a result that does not convert to ` +
+                        `an AuditResult: ${converted.detail}`,
                     work_item_id: entry.work_item_id,
                     result_path: entry.result_path,
                 });
                 continue;
             }
-            // Warnings are NOT rejections: an accepted result never refused anything,
-            // so a warning must never reach the rejection-classified issue list — the
-            // ONE ledger recorder would otherwise record kind:'rejected' for a result
-            // that was accepted, manufacturing a repair story that never happened.
-            // They ride the separate advisory channel instead (rendered for the
-            // operator; never counted as a submission that could not be accepted).
-            validation_warnings.push(...validationIssues
-                .filter((issue) => issue.severity === "warning")
-                .map((warning) => ({
+            // VALIDATE BEFORE ACCEPT. The conversion above proves only the envelope
+            // contract (`FindingSchema` admits an evidence-less finding); these are the
+            // per-result rules the downstream batch gate applies, applied HERE so an
+            // error-severity issue never reaches the accepted pair. A rejected item is
+            // simply never in the ledger, so the corrected file at the same bound path
+            // is re-read on the next fold — acceptance used to be terminal instead, and
+            // a failed batch gate then wedged the run permanently.
+            //
+            // Orphans (task pruned by a re-plan) pass through UNVALIDATED with the same
+            // stderr notice the batch gate uses — never newly rejected: refusing one
+            // here would strand it outside the append-only ledger entirely.
+            if (!activeTaskIds.has(entry.work_item_id)) {
+                process.stderr.write(`audit host-handoff ingest: result for '${entry.work_item_id}' is not in the ` +
+                    `active task manifest (orphaned by re-planning); retained in the accepted pair ` +
+                    `but skipped at the validation gate\n`);
+            }
+            else {
+                const validationIssues = validateOneAuditResult(converted.auditResult, [
+                    ...params.auditTasks,
+                ], {
+                    lineIndex: params.lineIndex,
+                });
+                const errors = validationIssues.filter((issue) => issue.severity === "error");
+                if (errors.length > 0) {
+                    issues.push({
+                        code: "result_validation_failed",
+                        check: "result_validation",
+                        message: `work item '${entry.work_item_id}' failed audit-results validation ` +
+                            `(${errors.length} error(s)); fix the result file at its bound path and call next-step again: ` +
+                            formatAuditResultIssues(errors),
+                        work_item_id: entry.work_item_id,
+                        result_path: entry.result_path,
+                    });
+                    continue;
+                }
+                // Warnings are NOT rejections: an accepted result never refused anything,
+                // so a warning must never reach the rejection-classified issue list — the
+                // ONE ledger recorder would otherwise record kind:'rejected' for a result
+                // that was accepted, manufacturing a repair story that never happened.
+                // They ride the separate advisory channel instead (rendered for the
+                // operator; never counted as a submission that could not be accepted).
+                validation_warnings.push(...validationIssues
+                    .filter((issue) => issue.severity === "warning")
+                    .map((warning) => ({
+                    work_item_id: entry.work_item_id,
+                    result_path: entry.result_path,
+                    message: `${warning.message} (${warning.field})`,
+                })));
+            }
+            // S7 quote-and-verify: the tool re-reads each cited span from disk and
+            // stamps the verdict. It NEVER rejects — a quote that does not re-verify
+            // rides through as `ungrounded` and synthesis surfaces it under "Ungrounded
+            // Findings (not confirmed)"; refusing here would discard the whole
+            // submission over one bad citation.
+            for (const finding of converted.auditResult.findings) {
+                finding.grounding = await verifyFindingGrounding(paths.root, finding, readSource);
+            }
+            resultIds.add(result.result_id);
+            // construction-site: AuditResult
+            // construction-site: Finding (the `findings` array; `lens` defaults from the enclosing contract, already spread in)
+            additions.push({
                 work_item_id: entry.work_item_id,
+                prompt_sha256: entry.prompt_sha256,
                 result_path: entry.result_path,
-                message: `${warning.message} (${warning.field})`,
-            })));
-        }
-        // S7 quote-and-verify: the tool re-reads each cited span from disk and
-        // stamps the verdict. It NEVER rejects — a quote that does not re-verify
-        // rides through as `ungrounded` and synthesis surfaces it under "Ungrounded
-        // Findings (not confirmed)"; refusing here would discard the whole
-        // submission over one bad citation.
-        for (const finding of converted.auditResult.findings) {
-            finding.grounding = await verifyFindingGrounding(paths.root, finding, readSource);
+                result_id: result.result_id,
+                result_sha256: contentSha256(result),
+                result,
+                audit_result: converted.auditResult,
+            });
         }
-        resultIds.add(result.result_id);
-        // construction-site: AuditResult
-        // construction-site: Finding (the `findings` array; `lens` defaults from the enclosing contract, already spread in)
-        additions.push({
-            work_item_id: entry.work_item_id,
-            prompt_sha256: entry.prompt_sha256,
-            result_path: entry.result_path,
-            result_id: result.result_id,
-            result_sha256: contentSha256(result),
-            result,
-            audit_result: converted.auditResult,
-        });
-    }
-    // The submission reading, contract checking and grounding re-verification
-    // above run UNLOCKED — they only read. The read-modify-write does not: the
-    // ledger is re-read under the lock and the additions are re-filtered against
-    // that fresh copy, so a concurrent writer's entries are merged rather than
-    // replaced, and a binding or result id it accepted in the meantime is not
-    // accepted a second time here.
-    let landed = [];
-    const ledger = await withAcceptedResultsLock(paths, params.logger, async (current) => {
-        const currentBindings = new Set(current.entries.map(bindingIdentity));
-        const currentResultIds = new Set(current.entries.map((entry) => entry.result_id));
-        landed = additions.filter((addition) => !currentBindings.has(bindingIdentity(addition)) &&
+        // Binding reads, validation and this accepted pair write share ONE hold.
+        // A concurrent prepare cannot replace the task bindings mid-ingestion.
+        const currentBindings = new Set(accepted.entries.map(bindingIdentity));
+        const currentResultIds = new Set(accepted.entries.map((entry) => entry.result_id));
+        const landed = additions.filter((addition) => !currentBindings.has(bindingIdentity(addition)) &&
             !currentResultIds.has(addition.result_id));
-        if (landed.length === 0)
-            return current;
-        const next = {
+        const ledger = {
             contract_version: ACCEPTED_RESULTS_CONTRACT_VERSION,
             run_id: params.runId,
-            entries: [...current.entries, ...landed].sort((left, right) => {
+            entries: [...accepted.entries, ...landed].sort((left, right) => {
                 const item = compareCodeUnits(left.work_item_id, right.work_item_id);
-                return item !== 0
-                    ? item
-                    : compareCodeUnits(left.prompt_sha256, right.prompt_sha256);
+                return item !== 0 ? item : compareCodeUnits(left.prompt_sha256, right.prompt_sha256);
             }),
         };
-        await writeAcceptedResults(paths, next);
-        return next;
+        if (landed.length > 0)
+            await writeAcceptedResults(paths, ledger);
+        for (const addition of additions) {
+            if (landed.includes(addition))
+                continue;
+            issues.push({
+                code: "duplicate_submission_id",
+                message: `work item '${addition.work_item_id}' was already accepted by a concurrent ` +
+                    `ingest of this run, so this submission was not accepted a second time`,
+                work_item_id: addition.work_item_id,
+                result_path: addition.result_path,
+            });
+        }
+        // Ledger recording is NOT done here. The shared submission ledger is written
+        // by the ONE recorder, `recordHostResultOutcomes`, at this ingest's only
+        // production caller — fed these very `issues` and `completed_work_item_ids` —
+        // so every rejection below already lands there in arrival order. A second
+        // writer inside the boundary would double-record the same fact.
+        const raw_issues = [...issues];
+        const refusals = await readTrailingSubmissionRefusals(paths.artifactsDir, raw_issues
+            .map((issue) => issue.work_item_id ?? issue.submission_id)
+            .filter((id) => id !== undefined));
+        const reportedIssues = enrichMissingSubmissionIssues(raw_issues, refusals, "submission_rejected");
+        await recordHostRootLogBoundary({ ...paths, phase: "ingest", logger: params.logger });
+        return {
+            accepted_count: landed.length,
+            accepted_results: ledger.entries.map((entry) => entry.audit_result),
+            accepted_results_path: paths.acceptedResultsPath,
+            validation_warnings,
+            completed_work_item_ids: [
+                ...new Set(ledger.entries.map((entry) => entry.work_item_id)),
+            ].sort(compareCodeUnits),
+            issues: reportedIssues,
+            raw_issues,
+        };
     });
-    for (const addition of additions) {
-        if (landed.includes(addition))
-            continue;
-        issues.push({
-            code: "duplicate_submission_id",
-            message: `work item '${addition.work_item_id}' was already accepted by a concurrent ` +
-                `ingest of this run, so this submission was not accepted a second time`,
-            work_item_id: addition.work_item_id,
-            result_path: addition.result_path,
-        });
-    }
-    // Ledger recording is NOT done here. The shared submission ledger is written
-    // by the ONE recorder, `recordHostResultOutcomes`, at this ingest's only
-    // production caller — fed these very `issues` and `completed_work_item_ids` —
-    // so every rejection below already lands there in arrival order. A second
-    // writer inside the boundary would double-record the same fact.
-    const raw_issues = [...issues];
-    const refusals = await readTrailingSubmissionRefusals(paths.artifactsDir, raw_issues
-        .map((issue) => issue.work_item_id ?? issue.submission_id)
-        .filter((id) => id !== undefined));
-    const reportedIssues = enrichMissingSubmissionIssues(raw_issues, refusals, "submission_rejected");
-    return {
-        accepted_count: landed.length,
-        accepted_results: ledger.entries.map((entry) => entry.audit_result),
-        accepted_results_path: paths.acceptedResultsPath,
-        validation_warnings,
-        completed_work_item_ids: [
-            ...new Set(ledger.entries.map((entry) => entry.work_item_id)),
-        ].sort(compareCodeUnits),
-        issues: reportedIssues,
-        raw_issues,
-    };
 }
 /**
  * Remove accepted entries — the supported way back out of an acceptance.