Meet app crash-looping: sandbox dies ~10s after startup with no error

Problem

The meet app is crash-looping on miren.garden. Sandboxes transition pending → running → dead in ~10 seconds with no error output in app logs or system journal.

Observations

• No error in app logs: The app starts up cleanly — DB initializes, Discord connects, server binds to port 3000, health check responds.
• No error in system journal: No OOM, no signal, no exit code logged. The sandbox just transitions to status.dead.
• Memory stable: Container stats show ~200MB, not growing.
• Timing: Consistently ~10 seconds from status.running to status.dead.
• Crash cooldown: After 7 consecutive crashes, hit cooldown: "application in crash cooldown until 2026-03-27T22:38:54Z (consecutive crashes: 7)"

Example lifecycle (sandbox CZUNbZX9hHNKY9HWvnRen)

22:27:58 status.pending → status.running
22:28:08 status.running → status.dead

Context

This started after deploying a config change to opt in to local storage via DB_PATH env var (commit 3810453). The DB itself is fine — exists at /miren/data/local/meet.db, 4096 bytes, WAL files present, directory is mounted.

Version

• App version: meet-vCZUNETRwfcYzxBsniW34t
• Pool: pool-CZUNEVnaCqgZp4ni8XoK2

Questions

• Is the process exiting cleanly (exit code 0) or being killed? The journal doesn't log exit codes for sandboxes.
• Could a health check timeout be killing it? No health check probe logs visible.
• Could the Bun process event loop be draining (no unhandled promise / no open handles keeping it alive)?