feat(producer): calibration-aware heartbeat + worker-death terminal-error contract

Field signals ts=1784019503 (heartbeat reports 0 frames during 64s
browser calibration — reads as broken but is healthy) and ts=1784042064
(1292s Windows render hard-exited during video frame extraction with
no final error string — silent worker crash).

Add calibrating/capturing state to heartbeat labels; surface synthetic
terminal error on unexpected worker exit when no explicit error was
emitted.

Stack: PR #6 of 9 (base via/overlay-count-lint).
Signed-off-by: Via <vance@heygen.com>

packages/engine/src/services/parallelCoordinator.test.ts

@@ -2,10 +2,12 @@ import { describe, it, expect } from "vitest";
 import {
   calculateOptimalWorkers,
   distributeFrames,
+  expectedFramesForTask,
   formatWorkerFailure,
   selectWorkerDiagnostics,
   shouldDisableBrowserPoolForParallelWorker,
   shouldVerifyWorkerGpu,
+  synthesizeSilentWorkerExitError,
   resolveParallelDeVerifySamples,
 } from "./parallelCoordinator.js";
 import type { EngineConfig } from "../config.js";
@@ -161,6 +163,91 @@ describe("worker failure diagnostics", () => {
       "Worker 1: Navigation timeout of 60000 ms exceeded; diagnostics: [FrameCapture:ERROR] page.goto failed mode=screenshot timeoutMs=60000",
     );
   });
+
+  // Field signal ts=1784042064: a Windows render hard-exited during video
+  // frame extraction without emitting a terminal error string; the parent
+  // treated the result as success because `error` was falsy. The synthesized
+  // string surfaces the shortfall so downstream telemetry can classify the
+  // failure instead of it disappearing silently.
+  it("synthesizes a terminal error when a worker exits with no error string", () => {
+    const message = formatWorkerFailure({
+      workerId: 3,
+      framesCaptured: 12,
+      startFrame: 0,
+      endFrame: 60,
+      durationMs: 180_000,
+    });
+    expect(message).toContain("Worker 3: worker 3 exited without terminal error string");
+    expect(message).toContain("framesCaptured=12");
+    expect(message).toContain("expected=60");
+    expect(message).toContain("range=[0, 60)");
+    expect(message).toContain("ts=1784042064");
+    expect(message).toContain("--workers=1");
+  });
+
+  it("prefers an explicit error string over the synthesized silent-exit message", () => {
+    const message = formatWorkerFailure({
+      workerId: 3,
+      framesCaptured: 12,
+      startFrame: 0,
+      endFrame: 60,
+      durationMs: 180_000,
+      error: "Target closed",
+    });
+    expect(message).toBe("Worker 3: Target closed");
+    expect(message).not.toContain("ts=1784042064");
+    expect(message).not.toContain("exited without terminal error string");
+  });
+
+  it("treats an empty error string as a silent exit for the message contract", () => {
+    const message = formatWorkerFailure({
+      workerId: 3,
+      framesCaptured: 12,
+      startFrame: 0,
+      endFrame: 60,
+      durationMs: 180_000,
+      error: "",
+    });
+    expect(message).toContain("exited without terminal error string");
+    expect(message).toContain("ts=1784042064");
+  });
+});
+
+describe("expectedFramesForTask", () => {
+  it("returns the range length for contiguous tasks", () => {
+    expect(expectedFramesForTask({ startFrame: 0, endFrame: 30 })).toBe(30);
+    expect(expectedFramesForTask({ startFrame: 10, endFrame: 25 })).toBe(15);
+  });
+
+  it("divides by stride for interleaved tasks", () => {
+    // Worker 0 in a 3-way interleave over 30 frames captures 0, 3, 6, ..., 27 → 10 frames.
+    expect(expectedFramesForTask({ startFrame: 0, endFrame: 30, frameStride: 3 })).toBe(10);
+  });
+
+  it("rounds up so the last-stride-remainder frame is expected", () => {
+    // Worker 0 in a 3-way interleave over 31 frames captures 0, 3, ..., 30 → 11 frames.
+    expect(expectedFramesForTask({ startFrame: 0, endFrame: 31, frameStride: 3 })).toBe(11);
+  });
+
+  it("returns zero for empty ranges", () => {
+    expect(expectedFramesForTask({ startFrame: 10, endFrame: 10 })).toBe(0);
+    expect(expectedFramesForTask({ startFrame: 30, endFrame: 10 })).toBe(0);
+  });
+});
+
+describe("synthesizeSilentWorkerExitError", () => {
+  it("names the field signal and reruns hint so operators can classify the failure", () => {
+    const message = synthesizeSilentWorkerExitError(
+      { workerId: 7, framesCaptured: 40, startFrame: 60, endFrame: 120 },
+      60,
+    );
+    expect(message).toContain("worker 7 exited without terminal error string");
+    expect(message).toContain("framesCaptured=40");
+    expect(message).toContain("expected=60");
+    expect(message).toContain("range=[60, 120)");
+    expect(message).toContain("Field signal ts=1784042064");
+    expect(message).toContain("--workers=1");
+  });
 });
 
 describe("shouldVerifyWorkerGpu", () => {

packages/engine/src/services/parallelCoordinator.ts

@@ -147,8 +147,49 @@ function compactDiagnosticLine(line: string): string {
   return line.replace(/\s+/g, " ").trim();
 }
 
+/**
+ * Expected frame count for a worker task, honoring its stride. Contiguous
+ * tasks (stride 1) expect `endFrame - startFrame`; interleaved tasks
+ * (stride > 1) expect `ceil((endFrame - startFrame) / stride)`, matching
+ * the loop shape in `captureFrameRange`.
+ */
+export function expectedFramesForTask(task: {
+  startFrame: number;
+  endFrame: number;
+  frameStride?: number;
+}): number {
+  const stride = task.frameStride ?? 1;
+  return Math.max(0, Math.ceil((task.endFrame - task.startFrame) / stride));
+}
+
+/**
+ * Synthetic terminal-error message for a worker whose exit didn't produce
+ * an explicit error string but under-captured its expected frame range.
+ * Field signal ts=1784042064: a 1292s Windows render hard-exited during
+ * capture with no final error string, leaving the operator with no
+ * actionable trace. This message surfaces the shortfall + reruns hint
+ * so downstream telemetry (and operators grepping logs) can classify the
+ * failure instead of it disappearing silently.
+ */
+export function synthesizeSilentWorkerExitError(
+  result: Pick<WorkerResult, "workerId" | "framesCaptured" | "startFrame" | "endFrame">,
+  expectedFrames: number,
+): string {
+  return (
+    `worker ${result.workerId} exited without terminal error string ` +
+    `(framesCaptured=${result.framesCaptured}, expected=${expectedFrames}, ` +
+    `range=[${result.startFrame}, ${result.endFrame})). ` +
+    `Field signal ts=1784042064 — this class of failure has been reported; ` +
+    `consider re-run with --workers=1 to isolate.`
+  );
+}
+
 export function formatWorkerFailure(result: WorkerResult): string {
-  const base = `Worker ${result.workerId}: ${result.error ?? "unknown error"}`;
+  const errorText =
+    result.error && result.error.length > 0
+      ? result.error
+      : synthesizeSilentWorkerExitError(result, expectedFramesForTask(result));
+  const base = `Worker ${result.workerId}: ${errorText}`;
   if (!result.diagnostics || result.diagnostics.length === 0) return base;
 
   const diagnostics = result.diagnostics.map(compactDiagnosticLine).join(" | ");
@@ -541,6 +582,17 @@ export async function executeParallelCapture(
     ),
   );
 
+  // A worker may return without an error string yet with framesCaptured
+  // below the task's expected count — that's the silent-exit shape field
+  // signal ts=1784042064 called out. Synthesize a terminal error string
+  // in-place so the filter below treats it as a failure (and so the
+  // caller's failure message actually names what went wrong).
+  for (const r of results) {
+    if (!r.error && r.framesCaptured < expectedFramesForTask(r)) {
+      r.error = synthesizeSilentWorkerExitError(r, expectedFramesForTask(r));
+    }
+  }
+
   const errors = results.filter((r) => r.error);
   if (errors.length > 0) {
     const errorMessages = errors.map(formatWorkerFailure).join("; ");