Has anyone seen Durable Object alarm handlers fail with cpuTimeMs: 0?

I'm getting intermittent exceptions (~<1%) where:
  • outcome: "exception" on alarm events
  • cpuTimeMs: 0 but wallTimeMs: ~30000 (30s timeout)
  • No console logs appear - code never executes
  • The error message in logs is just the scheduledTime timestamp, not an actual errorUsing SQLite-backed DOs (new_sqlite_classes). Seems like the DO fails to instantiate/cold start before any JS runs. The automatic alarm retry recovers it, but curious if anyone knows what causes this or how to reduce frequency.
Was this page helpful?