fix: prevent startup hang from blocking VLM call in redo recovery

[mvanhorn]

Description

Server startup can hang indefinitely when crash-recovery redo includes session_memory extraction. The LockManager.start() method awaits _recover_pending_redo() synchronously, which calls extract_long_term_memories() (a VLM call). If the VLM service is slow or unresponsive, the server never finishes starting and /health remains unavailable.

Two changes:

1. Move redo recovery to a background task via asyncio.create_task() so startup completes immediately
2. Add 60s timeout to extract_long_term_memories via asyncio.wait_for() to prevent indefinite hangs on individual redo tasks

Related Issue

Fixes #1222

Type of Change

• Bug fix (non-breaking change which fixes an issue)

Changes Made

• __init__: initialize self._redo_task
• start(): launch _recover_pending_redo() as background task instead of awaiting it
• stop(): cancel and await the redo task on shutdown
• _redo_session_memory(): wrap VLM call with asyncio.wait_for(timeout=60.0)

Testing

• Unit tests pass
• Platform testing (Linux)

Checklist

• My code follows the coding style of this project
• I have performed a self-review of my code
• My changes generate no new warnings

This contribution was developed with AI assistance (Codex).

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis ✅ 1222 - PR Code Verified Compliant requirements: Move redo recovery to background task to allow startup to complete immediately Add 60s timeout to extract_long_term_memories call Preserve fallback semantic enqueue behavior Requires further human verification: Platform testing (Linux) as mentioned in PR checklist

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🏅 Score: 90

🧪 No relevant tests

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ Recommended focus areas for review Missing loop check in redo task cancellation The redo task cancellation in stop() doesn't include the same loop check as the cleanup task, which could lead to errors if stop() is called from a different event loop. if self._redo_task: self._redo_task.cancel() try: await self._redo_task except asyncio.CancelledError: pass Unhandled exceptions in background redo task The background redo task doesn't have a done callback to log unhandled exceptions, which could result in silent failures of the redo recovery process. self._redo_task = asyncio.create_task(self._recover_pending_redo())

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
General	Add loop check for redo task cancellation Add the same event loop check for _redo_task as used for _cleanup_task to ensure consistent behavior when stopping the manager. This avoids potential issues if the task was created on a different event loop. openviking/storage/transaction/lock_manager.py [63-69] if self._redo_task: self._redo_task.cancel() try: - await self._redo_task + if self._redo_task.get_loop() is asyncio.get_running_loop(): + await self._redo_task except asyncio.CancelledError: pass self._redo_task = None Suggestion importance[1-10]: 5 __ Why: The suggestion makes _redo_task cancellation behavior consistent with _cleanup_task by adding an event loop check, which helps avoid potential issues if the task was created on a different loop. This is a useful maintainability improvement but not a critical fix.	Low

[mvanhorn]

Thanks for reviewing and merging!