Both Fiwano rows are checked every minute from four locations outside our own infrastructure. Downtime is recorded only when at least three of them agree, so a single failing probe never counts as an outage — and an interruption shorter than about two minutes may not register. The first row's check runs through the application and its database, which the API, message delivery and the portal all depend on; the second checks the public site, which renders without them. Delivery itself happens between Meta and your endpoint, where no outside probe can see, so a delivery problem reaches this page as a published incident rather than as a gap in a bar. The Meta rows repeat what Meta publishes, in Meta's own words; they are not our measurement and do not enter our figures.

We aim for 99.99% availability. The figure at the top and the bars cover the same 90 days, and the monthly table keeps every month since measurement started in September 2026; nothing earlier is published or backfilled. A failover counts as downtime whenever the probes saw it. Planned maintenance is excluded only when it was announced here at least 72 hours in advance. Published figures are never edited afterwards.

Every incident and maintenance window is a file in a public git repository, fiwano-com/fiwano-updates, and every update is a separate commit, so the record cannot be quietly improved afterwards. The incident history lists every outage the probes recorded, whether or not we wrote about it, next to what we published. When something is wrong, the headline shows when it started: the moment we published the incident or the moment the probes agreed the service was down, whichever came first. Times are shown in your browser's time zone. Machine-readable status is at /api/v2/summary.json (Statuspage v2 format) and /history.atom.