[BUG] Desktop (Windows, WSL): long-running asyncRewake SessionStart hook blocks new Code sessions (hang on first message)

Summary

After today's Claude Desktop auto-update on Windows, every new Claude Code session targeting WSL hangs on the first message. Sessions created before the update still work. In-app feedback also hangs (it attaches to the stuck session).

Environment

• Claude Desktop (Windows, MSIX) 2.42304.0, updated from 2.31226.0 on 2026-10-15
• Code engine 2.1.299 (was 2.1.293), session runs on wsl:Ubuntu-24.04
• Project has SessionStart / UserPromptSubmit hooks (all finish in < 1.2 s when timed directly)

Steps

1. Open Claude Desktop → new Code session in a WSL folder.
2. Type "hello".
3. Nothing happens. No transcript is ever written for the session.

Logs (main.log)

• [CCD CycleHealth] ... hadFirstResponse=false for every new session since the update (last good new session started before the update)
• Failed to get SDK commands for local_...: Error: Query closed before response received
• [CCD] initializationResult read failed for local_... (Error)
• Failed to toggle remote control for local_...: "Remote Control connect timed out" at each new session. It persists after turning Remote Control auto-enable off (rcAutoEnable verdict: enable=false), and the hang remains.

What I verified

• Running the same engine binary in the same folder with the desktop's exact argv and sending {"type":"control_request","subtype":"initialize"} over stdin returns control_response success in ~0.4 s; stdout closes cleanly within 5 s of stdin EOF.
• -p stream-json prompt in the same folder answers normally.
• So the engine responds; the response appears not to reach the desktop app for new sessions over the WSL remote-server bridge.
• Resumed (pre-update) sessions keep working on the same app/engine versions.

Expected

New sessions start and answer as before the update.