Monitor and troubleshoot
Integrations do not fail loudly. They fail at 02:00 in a job nobody watches, and the first symptom is a buyer asking why the invoice price is not the screen price. The worst failure of all is the one where nothing errors because nothing ran — which is why "no file arrived by 03:00" is an alarm you have to build deliberately.
Three things to keep up: a five-minute daily read of the four logs, a triage routine for the handful of failures that account for almost every support case, and somebody who reads the changelog so that a deprecation is a scheduled task rather than a surprise.
- Monitor your integrations — Calls, errors, latency, webhook delivery.
- Common API problems — 401, 403, 429, 422 — and the webhook that never arrives.
- Handling API changes — Deprecations, versions, the changelog.
Test an integration end to end
Fixtures, webhook replay, the idempotency tests most projects skip, and a go-live checklist you can actually hold an integrator to.
Monitor your integrations
The four logs worth reading, the three alerts worth setting, and the daily five minutes that catches an integration problem before a buyer does.