Status
System status.
Live operational state of the services we run. If something here is degraded, we already know and are on it.
- Updated as incidents happen
- No hidden downtime
- Every service listed
This page is maintained manually while automated monitoring is connected. If something looks wrong to you but reads green here, tell us — you are probably right and we are wrong.
Last 90 days
Uptime, unedited.
One bar per day. Hover any bar for the detail — degraded and failed days stay on the record rather than quietly rolling off.
Incident history
What went wrong, and why.
Every incident in the window above, written up the day it closed.
-
Resolved Dashboard unavailable for 41 minutes
A database migration ran during a period we had wrongly assumed was quiet and held a lock longer than expected. Sites and automations were unaffected; the dashboard returned 503 for the duration.
What changed: migrations now run behind a read-replica swap, and the maintenance window is confirmed against actual traffic rather than assumed.
-
Degraded Slow dashboard analytics
Following the incident above, the analytics view was rebuilding its cache and took up to nine seconds to load. Everything else responded normally.
What changed: the cache now warms before traffic is restored rather than on the first request after it.
-
Degraded Delayed workflow runs
A third-party API was rate-limiting us harder than its documented ceiling. Scheduled jobs ran up to 20 minutes late; none failed and none ran twice.
What changed: exponential backoff with jitter, plus an alert on queue depth rather than only on outright failure.
-
Degraded Slow image delivery in one region
An edge node served images from a distant origin for roughly two hours. Pages rendered correctly but images arrived late for some visitors in Asia.
What changed: origin failover is now regional instead of global.
Status