fix(webapp): set completedAt for FAILED synthetic run-header

Third surface of the same symmetric pattern (alongside
ApiRetrieveRunPresenter and buildSyntheticSpanRun). The run-detail
route derives `isCompleted` from `run.completedAt \!== null` and gates
SSE live-reloading on it (`route.tsx:459`, `:551`). Leaving the field
null for FAILED buffered runs would keep a terminal run live-reloading
indefinitely while the badge already says SYSTEM_FAILURE.

Falls back to `run.createdAt` for the FAILED case — same approximation
used by the sibling helpers. New unit test pins the contract for the
NavBar header.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: d-cs

apps/webapp/app/v3/mollifier/syntheticRunHeader.server.ts

@@ -50,7 +50,16 @@ export function buildSyntheticRunHeader(args: {
       : ("PENDING" as const),
     isFinished: isCancelled || isFailed,
     startedAt: null,
-    completedAt: run.cancelledAt ?? null,
+    // Symmetric with `buildSyntheticSpanRun` and the
+    // `ApiRetrieveRunPresenter` synth path. The run-detail route
+    // derives `isCompleted` from `completedAt !== null` and gates SSE
+    // live-reloading on it (`route.tsx:459`, `:551`); leaving
+    // `completedAt` null for FAILED would keep a terminal buffered run
+    // live-reloading forever. PG-resident SYSTEM_FAILURE rows always
+    // have completedAt set, so fall back to createdAt (the buffer
+    // entry has no separate failedAt — closest proxy for when the
+    // terminal state landed).
+    completedAt: run.cancelledAt ?? (isFailed ? run.createdAt : null),
     logsDeletedAt: null,
     rootTaskRun: null,
     parentTaskRun: null,

apps/webapp/test/mollifierSyntheticRunHeader.test.ts

@@ -86,6 +86,22 @@ describe("buildSyntheticRunHeader", () => {
     expect(header.completedAt).toEqual(CANCELLED_AT);
   });
 
+  it("populates completedAt for FAILED runs so the route stops live-reloading and renders as completed", () => {
+    // The run-detail route derives `isCompleted` from
+    // `run.completedAt !== null` and gates SSE live-reloading on it
+    // (`route.tsx:459`, `:551`). Leaving completedAt null for FAILED
+    // buffered runs would keep a terminal run live-reloading forever
+    // while the badge already says SYSTEM_FAILURE. Symmetric with
+    // buildSyntheticSpanRun + ApiRetrieveRunPresenter.
+    const header = buildSyntheticRunHeader({
+      run: makeSyntheticRun({ status: "FAILED" }),
+      environment: ENV,
+    });
+    expect(header.status).toBe("SYSTEM_FAILURE");
+    expect(header.isFinished).toBe(true);
+    expect(header.completedAt).toEqual(NOW);
+  });
+
   it("forwards identity and environment fields from the snapshot", () => {
     const header = buildSyntheticRunHeader({ run: makeSyntheticRun(), environment: ENV });
     expect(header.friendlyId).toBe("run_friendly_1");