fix: reconcile interrupted schedule executions - #6780

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		Aug 17, 2026 7:50pm

PR Summary

High Risk
Changes core schedule at-most-once recovery, occurrence claim guards, and async-job cleanup ownership—incorrect reconciliation could double-run or skip scheduled occurrences.

Overview
After outages, claimed schedule carrier jobs are reconciled from persisted workflow execution logs instead of being redispatched. Outcomes map to success, failure, cancellation, pause, or an indeterminate failure; only pending carriers with zero attempts still run inline.

The schedule tick adds recovery for stale processing, claimed-pending, and unreconciled terminal carriers, with claim-guarded applyScheduleSuccessUpdate / applyScheduleCancellationUpdate / failure accounting and metadata flags (scheduleReconciled, scheduleRecoveryIrrecoverable). Carrier lookup or enqueue uncertainty preserves the occurrence rather than deferring cadence via infra retry.

Stale execution cleanup terminalizes both running and redacting logs, excludes schedule-execution async jobs from generic stale pending/processing handling, and retains schedule carriers until reconciled (irrecoverable tombstones are not pruned).

^{Reviewed by Cursor Bugbot for commit c7a839a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Greptile Summary

This PR changes schedule outage recovery to reconcile already-claimed carriers from persisted workflow execution logs instead of redispatching them.

• Separates untouched pending database jobs from previously claimed jobs using the attempt count.
• Reconciles successful, failed, cancelled, paused, and indeterminate outcomes through occurrence-guarded schedule accounting.
• Preserves claims when queue lookup or enqueue acceptance is uncertain.
• Transfers active schedule-carrier cleanup to schedule recovery and retains terminal carriers until reconciliation is recorded.
• Extends stale workflow-log cleanup to interrupted redaction finalization.

Confidence Score: 5/5

The PR appears safe to merge; no concrete blocking or independently actionable non-blocking defect remains.

Claimed carriers are reconciled without redispatch, database settlement and schedule accounting are transactionally coordinated, claim guards prevent stale occurrences from advancing newer schedule state, and cleanup predicates preserve carriers until reconciliation completes.

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  Tick[Schedule cron tick] --> Claim[Claim due schedule occurrence]
  Claim --> Lookup[Look up deterministic carrier]
  Lookup -->|No carrier| Enqueue[Create and enqueue execution carrier]
  Lookup -->|Untouched pending, attempts = 0| Execute[Execute database fallback job]
  Lookup -->|Claimed or terminal carrier| Evidence[Read persisted workflow execution log]
  Evidence --> Classify{Classify outcome}
  Classify -->|Completed / paused| Success[Apply guarded success accounting]
  Classify -->|Cancelled| Cancel[Apply guarded cancellation accounting]
  Classify -->|Failed / indeterminate| Failure[Apply guarded failure accounting]
  Success --> Mark[Mark carrier reconciled]
  Cancel --> Mark
  Failure --> Mark
  Mark --> Cleanup[Retention cleanup may delete carrier]
  Enqueue -->|Acceptance uncertain| Preserve[Preserve occurrence claim for later reconciliation]

_{Reviews (1): Last reviewed commit: "simplify schedule recovery provider hand..." | Re-trigger Greptile}