Resilient failure recovery

Design UI/async code so it fails safely, cleans up deterministically, and recovers when the environment changes.

Apply these rules: 1) Gate irreversible transitions on success

• If a failure should leave the system usable, do not perform “detach/delete/sever” until the succeeding operation is confirmed.
• Example pattern:

  const { detach } = cloud;
  
  async function resume(sessionId: string) {
    try {
      // Resume local state first; only then detach cloud.
      await restoreLocalSession(sessionId);
      detach();
    } catch (e) {
      // On failure, keep cloud attached so user can retry.
      throw e;
    }
  }
  

2) Make cleanup robust to missing events

• Don’t assume pointerup, visibilitychange, etc. will always fire.
• Detect “stale drag” / “no button held” / “component unmounted” and run the same teardown path.

3) Implement explicit retry points tied to reconnection/environment changes

• For external-service lookups at startup, retry based on transport state changes (e.g., sidecar reconnect), not only on initial mount.
• Add a test for “first attempt fails, later reconnect succeeds.”

If you consistently (a) sequence side effects after success, (b) self-terminate when events don’t arrive, and (c) retry on reconnection, the codebase will handle real-world failure scenarios gracefully.