We mirror the public status pages of 17 providers every five minutes and keep every incident they post. The alert engine uses the live state to hold a page when the cause is upstream. This is the history behind that.
1,288 incidents kept since 5 Nov 2025. 17 open right now. Impact labels are the provider's own. Read the September report →
Azure: mirrored for alert decisions, but the feed only lists incidents while they're open and publishes no history, so there's nothing to keep yet.
SendGrid's status page is Twilio's page, so it's listed there.
Read the numbers with this in mind: each provider's feed gave us its 50 most recent incidents when we started keeping them, so the windows differ — Twilio's cover 59 days, Datadog's 369. “Median fix” and “< 1 hour” are over everything we hold; the 30-day count is comparable across providers from 10 Oct 2026 on. We don't rank providers by incident count until every window is a full quarter.
Data as JSON: /upstream.json, and per provider at /upstream/<provider>/incidents.json. CC BY 4.0.
The same data holds your pages when the cause is upstream. Probe quality →