method · every provider's API is probed each 5 min · a failed attempt is retried once before it counts · down = 2 consecutive failed cycles · recovery is announced
honesty · faults on our monitor's own network are discarded, never shown as provider downtime · providers without a probeable API show as unmonitored, never as up · gaps render as gaps
latency · response-time charts are measured on real requests from accounts that share telemetry (about a tenth of traffic) and include full streaming time — long agent turns are counted, so p95 reflects workload length, not queueing · days with few measured requests can swing widely (hover a point for its sample size) · probe rtt is API reachability only