Automation
Error handling in flows — what actually works
The default "run after" pattern is fine until it isn’t. Here is the escalation ladder we use.
Published by CloudFixies · 23 September 2025
Level 1 — retry
Built-in retry with exponential back-off. Solves 60% of transient failures.
Level 2 — dead-letter
Route unrecoverable failures to a dedicated queue with the full run context. A human triages them daily.
Level 3 — alert
If dead-letter volume exceeds a threshold, page the on-call engineer. Never rely on someone remembering to check the queue.
◆ Takeaway
Retry, dead-letter, alert. In that order. Not one of them alone.