Hosted status pages, incident updates and subscriber notifications by email, SMS, Slack and webhook. Used by 2,900 engineering teams.
Webhook retries queued behind a slow downstream endpoint. Queue drained; per-subscriber concurrency limits added so one slow receiver cannot delay the rest.
Elevated latency on the status API in eu-west-1 during a database failover. No updates were lost; subscriber notifications were delayed by up to four minutes.