flaky test - wait-for-host-standby: retry puppeteer.launch with diagnostics

[habdelra]

Why

Boxel CLI Tests on PR #4863 failed with this in the dev-stack startup:

[start:prerender-dev] [wait-for-host-standby] probing https://localhost:4200/_standby (max 600s)...
[start:prerender-dev] [wait-for-host-standby] unexpected failure: TimeoutError: Timed out after 30000 ms while waiting for the WS endpoint URL to appear in stdout!
    at ChromeLauncher.launch (puppeteer-core/src/node/BrowserLauncher.ts:260:15)
    at async main (packages/realm-server/scripts/wait-for-host-standby.ts:84:17)

puppeteer.launch timed out before the script's existing retry loop could even start. The retry loop only covered the post-launch page.goto + waitForFunction phases. A slow Chrome cold start on a loaded CI runner aborts the whole script, which fails the Start dev stack step, which fails the job — and Chrome cold-start time is a well-known flakiness vector on CI.

What this PR changes

packages/realm-server/scripts/wait-for-host-standby.ts: wrap puppeteer.launch in its own retry helper.

• explicit 90s launch timeout per attempt (default was 30s)
• up to 3 attempts with a 2s backoff between them, capped by the same 10-minute total deadline as the navigation loop
• structured log lines for each attempt (executable path, args, dumpio, timeout, success/failure timing) so a future flake is debuggable from the CI log alone
• on the final attempt, enable puppeteer's dumpio so Chrome's own stdout/stderr is piped through node — if launch still fails after the retries, we capture why (sandbox denial, missing shared library, GPU init crash) instead of a bare "Timed out … waiting for the WS endpoint URL"

The WAIT_FOR_HOST_STANDBY_VERBOSE=1 env var also now forces dumpio: true on every attempt for local repro.

What this does NOT do

• Doesn't touch BrowserManager in the prerender server itself. That path also uses the puppeteer default 30s launch timeout, but it runs under a long-lived service with its own restart/recovery story — out of scope for this PR.
• Doesn't change the post-launch retry budget. The 30s-per-phase + 10-minute total ceiling for page.goto/waitForFunction is unchanged.

Test plan

• CI passes on this branch
• Re-run of the failing job on PagePool: materialize the reserved module/command tab; plumb priority through definition lookups #4863 (after rebase / retarget) starts the dev stack cleanly

[Copilot]

Pull request overview

Wraps puppeteer.launch in wait-for-host-standby.ts with its own retry loop and a 90s per-attempt timeout to mitigate CI flakiness where Chrome cold-starts exceed Puppeteer's default 30s launch timeout before the existing post-launch retry loop ever runs.

Changes:

• Add launchBrowserWithRetry helper with up to 3 attempts, 2s backoff, and a 90s timeout per attempt, bounded by the existing 10-minute total deadline.
• Enable Puppeteer dumpio on the final attempt (and whenever WAIT_FOR_HOST_STANDBY_VERBOSE=1) to surface Chrome's own stderr when launch ultimately fails.
• Replace the single inline puppeteer.launch call in main() with the new retry helper.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

Host Test Results

    1 files      1 suites   1h 29m 31s ⏱️
2 661 tests 2 646 ✅ 15 💤 0 ❌
2 680 runs  2 665 ✅ 15 💤 0 ❌

Results for commit 4af133e.

Realm Server Test Results

    1 files      1 suites   8m 26s ⏱️
1 408 tests 1 407 ✅ 0 💤 1 ❌
2 990 runs  2 989 ✅ 0 💤 1 ❌

Results for commit 4af133e.

For more details on these errors, see this check.