I think down detector doesn't actually have any probes or actual insight into status etc. I think it uses search volume on its own service as a proxy for an outage - so if e.g. lots of people rush to down detector to query to see if SERVICE_FOO is down, it will register as an outage on down detector because loads of people are trying to see if there is an outage even if SERVICE_FOO is actually totally fine.
My hunch is everyone saw that openai and Claude were down and checked for Gemini too. I was using Gemini the whole time this happened without a blip so it certainly wasn't down in my region at least. 3.8 flash is pretty good and didn't miss a beat.
So Down Detector is a kind of quantum observation experiment, observing influences the result.
Down detector is a self-reported platform, they use a baseline over 6 months from user reports to try to automatically "detect" if there is a real outage, or if its just a couple end-users. Ultimately, the outage is entirely based on users going to down detector and clicking on "Report a Problem".
https://downdetector.com.py/en/methodology/