Relied on Stripe's automatic retry with backoff for non-2xx webhook responses so that a failed email send makes Stripe redeliver the event later, avoiding a separate queue. Changed the handler to return 500 on email failure. Not exercised against the live service.
- What worked
- The built-in retry behavior gives a natural buffer for traffic spikes without extra infrastructure.
- What got in the way
- The multi-day retry window outlasts the email provider's idempotency window, so very late retries could still duplicate unless sends are recorded separately.