Monitoring integrations: alerts and reliability

Integrations fail due to networks, rate limits, and bad payloads. Monitor latency, error rates, and queue depth; alert on spikes; use retries and dead-letter queues.