fix(engine): retry probe on pollHfReady zero-duration timeout

packages/engine/src/services/frameCapture-transientErrors.test.ts

@@ -15,6 +15,10 @@ describe("isTransientBrowserError", () => {
     "Failed to launch the browser process! TROUBLESHOOTING: https://pptr.dev/troubleshooting",
     "connect ECONNREFUSED 127.0.0.1:9222",
     "Navigation timeout of 60000 ms exceeded",
+    // pollHfReady timed out before window.__renderReady flipped true — the
+    // classic symptom of a slow/contended host (e.g. several renders running
+    // concurrently); a fresh browser session on retry usually clears it.
+    "[FrameCapture] Composition has zero duration.\n  Runtime ready: false, __player: true, __hf.seek: true, GSAP timeline: true, data-duration: 53.3s",
   ])("returns true for transient error: %s", (message) => {
     expect(isTransientBrowserError(new Error(message))).toBe(true);
   });
@@ -25,6 +29,10 @@ describe("isTransientBrowserError", () => {
     "Composition duration is 0",
     "SYSTEM_FONT_USED: -apple-system",
     "",
+    // The runtime finished initializing (renderReady: true) and still reports
+    // zero duration — a genuine authoring bug (no timeline, no data-duration),
+    // not a transient host hiccup. Must keep fast-failing without a retry.
+    "[FrameCapture] Composition has zero duration.\n  Runtime ready: true, __player: true, __hf.seek: true, GSAP timeline: false, data-duration: not set",
   ])("returns false for non-transient error: %s", (message) => {
     expect(isTransientBrowserError(new Error(message))).toBe(false);
   });

packages/engine/src/services/frameCapture.ts

@@ -1975,6 +1975,14 @@ const TRANSIENT_BROWSER_ERROR_PATTERNS = [
   /Failed to launch the browser process/i,
   /Navigation timeout of \d+ ms exceeded/i,
   /ECONNREFUSED/i,
+  // pollHfReady's own timeout — thrown when window.__renderReady never flips
+  // true within playerReadyTimeout. "Runtime ready: false" means init simply
+  // didn't finish in time (commonly a slow/contended host, e.g. several
+  // concurrent renders), which a fresh session usually clears on retry. This
+  // is distinct from the "Runtime ready: true" fast-fail case a few lines up
+  // in pollHfReady (no timeline + no data-duration) — that's a genuine
+  // authoring bug and intentionally NOT matched here, so it still fails fast.
+  /Composition has zero duration[\s\S]*Runtime ready: false/,
 ];
 
 export function isTransientBrowserError(error: unknown): boolean {

packages/producer/src/services/render/stages/probeStage.test.ts

@@ -69,6 +69,7 @@ mock.module("@hyperframes/engine", () => ({
   // live in frameCapture-transientErrors.test.ts — update both if patterns change.
   isTransientBrowserError: (error: unknown) => {
     const msg = error instanceof Error ? error.message : String(error);
+    if (/Composition has zero duration[\s\S]*Runtime ready: false/.test(msg)) return true;
     return /Navigating frame was detached|Target closed|Session closed|browser has disconnected|Page crashed|Execution context was destroyed|Cannot find context with specified id|Failed to launch the browser process|Navigation timeout of \d+ ms exceeded|ECONNREFUSED/i.test(
       msg,
     );
@@ -337,6 +338,49 @@ describe("runProbeStage — transient browser error retry (#1687)", () => {
     expect(closeCaptureSessionCallCount).toBe(2);
   });
 
+  it("retries once on a pollHfReady zero-duration timeout (renderReady: false) and succeeds", async () => {
+    resetRetryMocks();
+    capturedCfgs.length = 0;
+    initializeSessionError = new Error(
+      "[FrameCapture] Composition has zero duration.\n  Runtime ready: false, __player: true, __hf.seek: true, GSAP timeline: true, data-duration: 53.3s",
+    );
+    initializeSessionFailUntilAttempt = 1;
+
+    const { runProbeStage } = await import("./probeStage.js");
+    const input = makeProbeInput({ cfgForceScreenshot: false, stageForceScreenshot: false });
+
+    const result = await runProbeStage(input);
+
+    expect(initializeSessionCallCount).toBe(2);
+    expect(closeCaptureSessionCallCount).toBe(1);
+    expect(result.duration).toBe(5);
+    expect(result.probeSession).not.toBeNull();
+  });
+
+  it("throws immediately on a permanent zero-duration error (renderReady: true — genuine authoring bug)", async () => {
+    resetRetryMocks();
+    capturedCfgs.length = 0;
+    initializeSessionError = new Error(
+      "[FrameCapture] Composition has zero duration.\n  Runtime ready: true, __player: true, __hf.seek: true, GSAP timeline: false, data-duration: not set",
+    );
+    initializeSessionFailUntilAttempt = 999;
+
+    const { runProbeStage } = await import("./probeStage.js");
+    const input = makeProbeInput({ cfgForceScreenshot: false, stageForceScreenshot: false });
+
+    let caught: unknown;
+    try {
+      await runProbeStage(input);
+    } catch (err) {
+      caught = err;
+    }
+
+    expect(caught).toBeInstanceOf(Error);
+    expect((caught as Error).message).toContain("Runtime ready: true");
+    expect(initializeSessionCallCount).toBe(1);
+    expect(closeCaptureSessionCallCount).toBe(1);
+  });
+
   it("retries on a transient browser LAUNCH failure (createCaptureSession throws)", async () => {
     resetRetryMocks();
     capturedCfgs.length = 0;