api.limitguard.ai shown on status.limitguard.ai since the service was restored on 2026-09-05. The longer outage before that has its own post-mortem.
Planned maintenance, 2026-09-19
All times UTC.
About 50 seconds of downtime. Requests in that window could not connect. Every service restarted from its existing data. The upgrade was run in the lowest-traffic hour.
Unplanned downtime, 2026-09-08 to 2026-09-18
Daily totals as measured by the status-page monitors. The monitors report per-day totals, not individual start times.
The causes of these windows are still being investigated, and none of them has been ruled a false positive yet. This page will be updated with a cause per row. Any that turns out to be a real outage over a few minutes will also get its own entry on the status page, whose incident history currently shows none of them.
Candidate contributor, added 2026-09-23: every deploy to
main stops the running
API before the new one listens, producing an empty-body 503 for roughly 43–45 s
(measured in infra/contabo/README.md, “Deploy window (#430)”). 119 commits landed
on main in this window, but they do not line up with the daily totals: 2026-09-08
had the most (over 75) and only 3 minutes of API downtime, while 2026-09-13 had none
and 7 minutes. Deploys can explain part of some rows at most, not the pattern as a
whole — this has not been reconciled per-row yet.