Research and incident write-ups

Two kinds of page live here. A write-up is the measured record of one outage we detected: which of our regions failed first, which recovered last, what people reported and what the vendor's own status page claimed over the same hours, frozen at publication. A reliability report ranks services within a category over a fixed window. The methodology page is the single description of what a reading means. Read as a feed: Atom.

Nothing published yet. The first pieces will be rapid write-ups of outages our probes detect, published within hours of the event with the regional timeline and nothing invented; reliability reports follow once the services they rank have a full quarter of coverage. The schedule below says when.

Recomputed on each visit from probe-detected outage events only, for services we watched throughout the window. A category shows a ranking only once every service in it has the coverage the claim needs; until then it says so, with the date that changes.

No active services to rank yet.

Claims scale with coverage age. Each kind of piece has a coverage bar every service it speaks for must clear, and a sample floor. The dates below are computed from when our record actually began; a ranking of younger services unlocks later, per its youngest member. Methodology in force: version v2-2026-08-22.

The sample floor counts only what has a full record over the window; a service added part-way through is left out of a ranking, by name, rather than ranked on part of it.