diff --git a/tests/runner-e2e/QUALIFICATION-2026-09-21.md b/tests/runner-e2e/QUALIFICATION-2026-09-21.md
index ba83b4b706..154ff4f289 100644
--- a/tests/runner-e2e/QUALIFICATION-2026-09-21.md
+++ b/tests/runner-e2e/QUALIFICATION-2026-09-21.md
@@ -115,3 +115,66 @@ This qualifies these two-turn grounding stories, not general answer quality,
 statistical reliability, multilingual behavior, or arbitrary long conversations.
 The five review dimensions remain factual grounding, stale-premise correction,
 honest uncertainty, useful next step, and clear prose. No production prompt changed.
+
+## Active reassignment: corrected live proof
+
+[Campaign 35659014397](https://github.com/paperclipai/paperclip/actions/runs/35659014397)
+on `cf6d4ae3a8576d822861bc916e7f536d0a0b7bc7`: **2/2 passed**, both
+cleanup passes, using `gpt-5.6-sol` and `claude-sonnet-5`.
+
+The boundary snapshot proves the original worker was running with a saved draft.
+Chat then reassigns the existing task to a second agent. The original stops with
+`issue_reassigned` before the successor starts; exactly one successor completes
+the same task. The plan, scope, and original draft revision survive. The successor
+may revise the canonical document, and its final contribution must be attributed
+to that successor. The audit records the native reassignment tool.
+
+The preceding [campaign 35657945095](https://github.com/paperclipai/paperclip/actions/runs/35657945095)
+on `6039b02ed` remains failed: both handoffs actually completed, but the draft
+oracle incorrectly required the latest document to remain frozen. Both successors
+legitimately revised that document. The new oracle requires the exact original
+revision to remain retrievable through the history API while allowing progress.
+Negative calibration rejects deletion or alteration of that revision. Both crash
+cases in the preceding campaign independently reproduced cleanup quarantine.
+
+## Measurement limits
+
+All selected environments here are local Linux GitHub Actions workers, not
+Daytona-hosted execution or the user's staging company. The result JSON retains
+source SHA, suite definition hash, profile/model, attempt, timing, token usage,
+and billing coverage; initial failures are never regraded as passes.
+
+Onboarding reports complete billing for only 4/26 results; active handoff reports
+incomplete billing for both interrupted workers; the two answer-quality results
+report complete coverage. Provider-reported monetary cost is zero while billing
+type is unknown, and local runtime is not metered. This does **not** establish
+zero actual spend. Recorded total tokens (including cached input) are 9,812,045
+for onboarding, 825,131 for final handoffs, and 1,273,098 for grounded answers.
+
+Public campaign viewers use the standard
+[Product E2E history](https://d1p6rlowie26tp.cloudfront.net/runner-e2e/)
+with the `gha-<workflow-run-id>-<attempt>` campaign identifier. These are bounded
+story qualifications, not a claim that all native-runner reliability is solved.
+
+## Crash guard follow-up
+
+[Campaign 35658772755](https://github.com/paperclipai/paperclip/actions/runs/35658772755)
+did not reach provider cases: GitHub artifact finalization returned HTTP 403.
+A replacement [campaign 35659580100](https://github.com/paperclipai/paperclip/actions/runs/35659580100)
+on `a11bd236e33833dc081cc3702baa3d3f98d8d12f` retained two failed recovery
+attempts, both with successful disposable cleanup. The API correctly refused
+Retry with 409 and created no second run, but the eval then waited for a run
+that had never been admitted.
+
+Retained network evidence explains the transition: `provider_transport_failed`
+first schedules a same-run retry; that attempt then reaches cleanup quarantine.
+The fixture now waits for this recovery classification before requesting a user
+retry. Screenshots also exposed a separate Retry button in the task-recovery
+banner. That banner previously recognized only native continuation reconciliation;
+it now also recognizes cleanup quarantine and links to Inspect run. Its regression
+test failed before the fix and passes afterward. The live guard must verify this
+link is rendered, no Retry remains, the API refuses retry, and saved work survives.
+
+These corrections do not supply a successful post-crash continuation. Worker-crash
+recovery remains unqualified until an agreed recovery policy is implemented and
+its successful outcome passes the eval.
diff --git a/tests/runner-e2e/catalog.ts b/tests/runner-e2e/catalog.ts
index f53649fae2..4f39bec921 100644
--- a/tests/runner-e2e/catalog.ts
+++ b/tests/runner-e2e/catalog.ts
@@ -975,7 +975,7 @@ export const runnerSuites: readonly RunnerSuiteFixture[] = [
     profiles: runnerProfiles.filter(profile => ["runner-codex", "runner-acpx-claude"].includes(profile.id))
       .map(profile => productionStoryProfile(defaultPermissionProfile(profile))),
     environments: [localEnvironment], tasks: chatQualificationTasks, expectedMatrixSize: 6,
-    definitionMetadata: { version: 5, permissions: "production-defaults", instructions: "production", crashBoundary: "verified-native-worker-pid-at-file-wait", recovery: "user-visible-retry", answerGrading: "exact-grounded-propositions-plus-separate-semantic-review", scheduling: "explicit-only" },
+    definitionMetadata: { version: 6, permissions: "production-defaults", instructions: "production", crashBoundary: "verified-native-worker-pid-at-file-wait", recovery: "user-visible-retry", answerGrading: "exact-grounded-propositions-plus-separate-semantic-review", scheduling: "explicit-only" },
   },
   ...(process.env.PAPERCLIP_RUNNER_E2E_CONNECTION_REVIEWS === "1" ? [connectionReviewSuite] : []),
   {
diff --git a/tests/runner-e2e/chat-qualification.ts b/tests/runner-e2e/chat-qualification.ts
index 7bf5610c6b..102016ce82 100644
--- a/tests/runner-e2e/chat-qualification.ts
+++ b/tests/runner-e2e/chat-qualification.ts
@@ -151,7 +151,14 @@ export async function runWorkerCrash(context: Context) {
     await input.evidence("chat-worker-fault-delivered.json", fault);
     context.expectedStops.set(boundary.id, "failed");
     await expect.poll(async () => (await input.api.get<ChatRun>(`/api/heartbeat-runs/${boundary.id}`)).status, { timeout: 120_000 }).toBe("failed");
-    const failed = await input.api.get<Row>(`/api/heartbeat-runs/${boundary.id}`);
+    // A failed transport may still have a scheduled same-run retry. Wait for
+    // recovery classification before treating it as available for a user Retry.
+    let failed: Row = {};
+    await expect.poll(async () => {
+      failed = await input.api.get<Row>(`/api/heartbeat-runs/${boundary.id}`);
+      return ["failed", "recovery_needed"].includes(failed.execution?.phase);
+    }, { timeout: 120_000 }).toBe(true);
+    await input.evidence("chat-worker-settled-failure.json", failed);
     await input.capture("worker-failed", "Worker loss before user Retry", "worker-failed.png");
     await writeFile(wait.gate, reference);
     if (failed.errorCode === "native_session_cleanup_quarantined") {
@@ -159,6 +166,7 @@ export async function runWorkerCrash(context: Context) {
       // the UI/API must not offer an attempt which cannot pass cleanup admission.
       await input.page.reload({ waitUntil: "domcontentloaded" });
       await expect(input.page.getByTestId("task-chat-composer-input")).toBeVisible();
+      await expect(input.page.getByRole("status", { name: "Task recovery" }).getByRole("link", { name: "Inspect run" })).toBeVisible();
       await expect(input.page.getByRole("button", { name: "Retry", exact: true })).toHaveCount(0);
       const refused = await input.api.request.post(`/api/agents/${input.fixtures.agent.id}/wakeup`, {
         data: { failedRunId: failed.id, reason: "retry_failed_run" },
@@ -173,7 +181,13 @@ export async function runWorkerCrash(context: Context) {
     }
     const retry = input.page.getByRole("button", { name: "Retry", exact: true });
     await expect(retry).toHaveCount(1, { timeout: 60_000 });
-    await retry.click();
+    const [retryResponse] = await Promise.all([
+      input.page.waitForResponse(response => response.request().method() === "POST" &&
+        new URL(response.url()).pathname === `/api/agents/${input.fixtures.agent.id}/wakeup`),
+      retry.click(),
+    ]);
+    await input.evidence("chat-worker-retry-response.json", { status: retryResponse.status(), body: await retryResponse.json() });
+    expect(retryResponse.ok(), "The visible Retry must admit an attempt before waiting for its result").toBe(true);
     await context.idle(2);
     const e = { boundary, failed, runs: await context.allRuns(), issueId: context.issue().id,
       prompt, comments: await context.comments(), reference, marker, planBefore,
diff --git a/ui/src/components/ExecutionBlockerNotice.test.tsx b/ui/src/components/ExecutionBlockerNotice.test.tsx
index 45c79bf0b4..3e8d42b420 100644
--- a/ui/src/components/ExecutionBlockerNotice.test.tsx
+++ b/ui/src/components/ExecutionBlockerNotice.test.tsx
@@ -49,11 +49,11 @@ describe("stopped task recovery notice", () => {
     expect(container.textContent).toContain("Verify the external action outcome before continuing.");
     expect(container.textContent).not.toContain("Automatic recovery of this task stopped.");
   });
-  it("links to the source run instead of offering a retry rejected by native reconciliation", async () => {
+  it.each(["native_continuation_requires_reconciliation", "native_session_cleanup_quarantined"])("links to the source run instead of offering a retry rejected by %s", async (cause) => {
     await act(async () => root.render(<QueryClientProvider client={client}>
       <ExecutionBlockerNotice companyId="company" issueId="task" onRetried={onRetried} blocker={{
         recoveryActionId: "recovery", runId: "failed-run", agentId: "agent",
-        cause: "native_continuation_requires_reconciliation",
+        cause,
         nextAction: "Inspect the original failure and reconcile the previous execution before continuing.",
       }} />
     </QueryClientProvider>));
diff --git a/ui/src/components/ExecutionBlockerNotice.tsx b/ui/src/components/ExecutionBlockerNotice.tsx
index dd8356059e..527ef3c6c8 100644
--- a/ui/src/components/ExecutionBlockerNotice.tsx
+++ b/ui/src/components/ExecutionBlockerNotice.tsx
@@ -19,7 +19,8 @@ export function ExecutionBlockerNotice({ companyId, issueId, blocker, onRetried
   });
   const failedRun = runs?.find(run => run.runId === blocker.runId &&
     ["failed", "timed_out"].includes(run.status));
-  const requiresInspection = blocker.cause === "native_continuation_requires_reconciliation";
+  const requiresInspection = blocker.cause === "native_continuation_requires_reconciliation" ||
+    blocker.cause === "native_session_cleanup_quarantined";
   const retry = useMutation({
     mutationFn: () => agentsApi.retryFailedRun(failedRun!.agentId, failedRun!.runId, companyId),
     onSuccess: () => {