Cloudflare Incident History

What these numbers are not

• A product only exists on the status page from the day Cloudflare created its component, so every chart starts each product at that date. Without that cut, R2 and D1 would appear to have had a flawless first decade.
• Roughly half of all incidents are tagged against no product at all. Some are network or data centre events, some name only a location, and some name nothing. They count towards the totals and towards no product.
• Because of that, product names are also read out of incident titles: "Elevated errors for Durable Objects in WNAM region" carries no tag but is plainly a Durable Objects incident. That recovers a few hundred, and Cloudflare frequently tags an incident Dashboard when the title calls it a Pages or Workers problem. Only titles are read, never the update text, because retros mention products in passing and that pins network-wide events onto whatever they happened to name. Switch Attribution to "Tags only" to see Cloudflare's tagging alone.
• 8 products rated here have no component of their own, so they get no chart. Rate limiting is reported inside Firewall, and Cloudflare publishes nothing separate for Pipelines or Secrets Store.
• One incident can name several products, so per-product counts add up to more than the incident total.
• “Time open” is how long the status page entry stayed open, not how long anything was actually broken, and concurrent incidents are counted separately. A team that posts early and closes late looks worse than one that posts nothing.
• Severity is Cloudflare's own impact field. I have not re-graded anything. Incident counts are a measure of how often Cloudflare tells you something is wrong, which is not the same as how often something is wrong.

Source: the JSON API behind cloudflarestatus.com. Snapshot taken August 21, 2026 and refreshed weekly. Not affiliated with Cloudflare.