Timeouts should restart hung Python subprocess/worker

[bbopen]

Summary

JS-side timeouts only reject the pending call; the Python process/worker keeps running, which can wedge the bridge and cause cascading timeouts or late-response protocol issues.

Failure mode

• Long-running call times out; subsequent calls continue to hit the same hung process.
• Late responses can still be parsed and cause protocol errors once timeout tracking expires.
• Mixed concurrency: a slow call can starve a fast call on the same subprocess, so the fast call times out even though it would complete quickly (serial bridge limitation).

Proposed fix

• On timeout, terminate and respawn the Python process (NodeBridge) or worker (OptimizedNodeBridge).
• Clear pending/timed-out IDs for the terminated process.
• Consider per-request cancellation protocol if feasible.

Acceptance criteria

• A call that sleeps beyond timeout triggers a process/worker restart.
• Next call after timeout succeeds without protocol errors.
• Integration test uses a slow Python function and asserts recovery.
• Concurrent test: one slow call and one fast call; slow call times out, fast call still succeeds (after restart or via separate worker).

[bbopen]

✅ Resolved in PR #136: refactor: unify NodeBridge and OptimizedNodeBridge (ADR-0001)

The unified NodeBridge now quarantines and restarts workers on timeout:

// BridgeCore callbacks trigger quarantine
onFatalError: (error: Error): void => this.quarantineWorker(worker, error),
onTimeout: (error: Error): void => this.quarantineWorker(worker, error),

When quarantineWorker is called:

1. Worker is marked as quarantined (excluded from selection)
2. Worker is force-terminated
3. A replacement worker is spawned if below minProcesses

This ensures hung processes are killed and the bridge recovers automatically.