Scheduled job stuck in PENDING for over a day; Retry fails with "failed to return job" (repo had two kernel-out-of-memory failures)

[takesoo]

What would you like help with?

None

Please share the URL to the organisation you're creating this request for

https://github.com/canbright

What happened

Our repository stopped receiving any Renovate activity after two consecutive
kernel-out-of-memory job failures. The weekly scheduled run did not happen
this week and the scheduled job has been stuck in PENDING for over a day.

Timeline (UTC):

• 2026-06-29 04:13 — job b33b4454-d434-4084-97aa-3e08cc21aa80 FAILED with
  kernel-out-of-memory. Duration 3m4s (of 30m limit), Memory Used 3.0GB (of 3.0GB limit).
  The debug log ends abruptly while running
  pnpm install --lockfile-only --recursive --ignore-scripts --ignore-pnpmfile
  (full cold-cache lockfile re-resolution of a pnpm monorepo).
• 2026-06-29 04:16 — job c117a2f3-4662-48b3-b81a-80079ca710a0 FAILED the same way
  (3m30s, 3.0GB/3.0GB).
• Since then, no job has completed for this repository. Our Dependency
  Dashboard issue was last updated 2026-06-29 04:10 UTC.
• A job with Reason=Scheduled (019f2fae-c591-713a-b582-b9abc1b39c8a) has been
  stuck in PENDING ("Job is pending and has not been run yet") since before
  our weekly schedule window (Monday 2026-07-06, schedule: before 9am on monday,
  timezone Asia/Tokyo). As of 2026-07-07 it still has not run, so no PRs were
  created this week.

What we tried

• Clicking Retry on the failed jobs in the developer portal → an error toast
  saying "failed to return job", and no new job appears.
• Checking multiple update checkboxes on the Dependency Dashboard issue,
  including the "Check this box to trigger a request for Renovate to run again
  on this repository" manual-run checkbox (2026-07-07 04:41 UTC) → no job was
  enqueued; the checkboxes remain checked and no branches/PRs were created.
• The Dependency Dashboard body has not been rewritten since 2026-06-29 04:10 UTC:
  it still lists PRs Refactor worker for testability #99/Update dependency gl-got to version 6.0.1 #100/Update dependency gl-got to version 6.0.2 #101 as open even though they were merged on
  2026-06-30. This confirms no Renovate job has run against this repository
  since the two kernel-out-of-memory failures.

Questions

1. Is this repository being throttled or quarantined on the Mend side after the
   consecutive kernel-out-of-memory failures? If so, can it be re-enabled?
2. Can you unstick / re-run the PENDING scheduled job (019f2fae-c591-713a-b582-b9abc1b39c8a)?
3. Is the "failed to return job" error on Retry a known issue?
4. The OOM itself: the 3.0GB memory limit is hit during pnpm's full lockfile
   re-resolution. Is there any recommended configuration to reduce memory usage
   on the hosted runners, given that resource increases appear to be limited to
   OSS-licensed projects?