Cloudflare logged a short incident on 28 September 2026 affecting Replicate Flux hotswap models, lasting 19 minutes. The fault sat on Cloudflare's side of the system, and it was brief enough that it may not have registered on your own checks.
Started
20:11 UTC
Duration
Lasted 19m
Source
IsDown
Next time
Cloudflare is degraded. Is your product working?
An incident on their side does not always mean an outage on yours, and the only way to know is to already be watching your own endpoints. Set that up in under a minute and you will be notified the next time your own site is affected. And there will be a next time.
50 monitors for your own websites. No card.
What happened?
Cloudflare's status page shows one incident starting at 20:11 UTC and running for 19 minutes on 28 September 2026. The title given is "Replicate Flux Hotswap Models Stuck", and the single update filed says Cloudflare was investigating Flux hotswap models that hadn't made progress since the previous Thursday, 24 September 2026. That phrasing suggests the underlying stall had been building for days before Cloudflare opened a public incident for it. Only one update was ever posted, marked "investigating", and no follow-up resolving or identifying message is on record. No components are listed as affected in Cloudflare's own breakdown, and no user reports came in during the window. With a single filing and no resolution note, this page can't tell you whether the stuck models cleared on their own or were fixed and simply never written up. Cloudflare's 90 day history shows 195 incidents with a median duration of 1 hour 47 minutes, so a 19 minute incident sits well below that typical length. Across all time, Cloudflare has logged 3,398 incidents. None of that volume tells you anything extra about this specific one beyond what the single update says.
Learning
UptimeRobot's own probes against Cloudflare's api-control, dns-1111 and edge-data components in the EU and NA regions ran at 100% uptime across the 24 hour, 7 day and 30 day windows, and this incident was not registered against any of them. That's not a contradiction. The stuck Flux hotswap models were a specific workload issue inside Cloudflare's platform, not a failure of the edge, DNS or API control paths that external checks poll every 60 seconds. A general availability probe watches whether the front door opens, not whether every job type behind it is making progress. This is the shape to expect from provider-side incidents that sit deeper in a vendor's stack than the surface endpoints: your pings stay green while a narrower feature quietly stalls for someone. Your own monitoring covers your half of the system, the parts you run and configure. UptimeRobot watching Cloudflare's own status feed and probe results covers the other half, so next time something like this happens you know which side broke without guessing.
