Keep slow dev renderer startup recoverable

[augmentcode]

Summary

The five-second development timeout rejected initializeTab before it could send wave-init. A late renderer ready IPC still resolved the original handshake, but startup had already been abandoned, leaving a hidden/blank renderer.

• Keep five seconds as a one-shot diagnostic threshold for both initPromise and waveReadyPromise; reveal the window/DevTools and continue waiting instead of treating slowness as fatal.
• Target DevTools at the awaited renderer, preserve promise rejection propagation, and abort development waits on window closure, renderer destruction, or renderer-process loss.
• Remove timers/listeners on completion or failure. Production waiting and IPC/hot-spare behavior are unchanged.

A larger fixed timeout would only move the race. A live renderer that never signals readiness still remains pending with diagnostics available; automatic crash recovery and production timeouts are outside this fix.

The final diff is one production-source file, +39/-20 lines. The oversized 212-line mocked Electron test harness was removed following review; it was disproportionate to this change. The main-process fix is byte-for-byte unchanged from the version verified in Electron.

Validation

• Actual task electron:linuxquickdev, as codespace, with isolated app data and sandbox intact: naturally crossed the diagnostic threshold at 19.923s, sent normal wave-init at 20.128s, and completed the wave-ready wait at 20.332s. Vite was ready at 13.840s; mapped-window timing was not treated as app readiness.
• Inspected rendered app pixels and an integrated terminal executing printf 'WAVE_RUNTIME_OK\n', producing the marker and returning a prompt. No runtime delay injection was needed.
• task check:ts: 17 existing errors in three preview fixture files. A read-only compiler comparison against the exact unchanged main-process source produced identical diagnostics; the fix adds none.
• git diff --check: passed, including after removal of the test harness. Confirmed the runtime-verified main-process source did not change in that follow-up.
• The initial 14 mocked test cases passed during implementation, but are not part of the final PR. Electron was not rerun merely for deleting that test-only file.
• Task-generated lockfile metadata churn excluded. No backend/wsh rebuild or dependency upgrade. Task processes and private scratch logs/data cleaned up.

Original investigation, tradeoffs, checks, and timing report · Rendered app screenshot · Terminal verification screenshot

Cosmos evidence links require VFS permissions. The linked report is a historical record of the original local-validation stage, including the test harness subsequently removed; this PR description reflects the final scope. Raw runtime logs are not uploaded. The upstream first-load delay remains unproven. Other operating systems, production packaging, and the full test suite were not run.

[coderabbitai]

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Advanced

Run ID: 39940a81-c248-478a-b53f-03b10279289d

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Autopilot is currently an internal CodeRabbit preview.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}