As I usually do, I will open up the problem a bit more, almost a fun story about 3-5 points of failure.
Not that fun after 4h of investigations around midnight. Some of this is probably technically incorrect but it’s a more fun read with possible errors.
Base facts (sorry explanations): My sso redirects to my home server (after Gandi failed to solve some things a few years back and I moved to Dynadot). Also works better redirecting after Hostpapa bough the very reliable LFC hosting and you could not be sure about them upgrading something silently and your whole infra coming down.
Now for the 3 (or could I even count this as a 5 point failure, would be a new record)
1*) Opera updated today, again, apparently adding some code that really (dis)likes 301/302 redirections and hard caches them (Chromium something).
2*) Dynadot apparently had a 30+ second lag tonight (which is what is needed to trigger the Chromium problem, which now got cached, uh oh).
3*) Discourse crashed (the oops mentioned above) because it could not access sso within 30 seconds. App had to be relaunched but I’m actually not sure if Discourse was down until relaunch or not (I was in panic mode 
). Gemini had me running in loops around sso/rails/redis etc based on the logs though. This is after I tried to upgrade but mentioning it under this point to keep this coherent.
4*) As I cannot figure out whats going wrong, I reboot the linux just in case and start to upgrade Discourse. Which fails because of the github problem. And yes, had to clean disks a bit first as usual but I was never at 0G which would have been a possible reason as well for redis problems. And yes II, now that things seem stable I will add 10G in the cloud and maybe Arkshine upgrade.
5*) Kinda the 1st problem but while trying to solve all this I continiously hard refresh the page to see if something would have been fixed somewhere. I’m guessing I’m hitting the cached redirect from the initial failure all the time though, which means I cannot even see if the problem has been fixed, even after the first relaunch.
- I can access my sso just fine.
- Another domain site not using sso also works well.
- Disabling sso for discourse, shows the forum, but I have never had normal logins work so not much help there. But at least I now know it really is sso messing things up.
- Gemini suggests testing in a private window so I actually fire up Firefox instead and, what, the forum works just fine. Great, except… whaaaat?
- Opera still fails. Yes, still caching the redirect I suspect but not something I knew at the time. Not going to clear site cookies, caches etc yet.
- Since things work I did the necessary work I actually intended to do using Firefox for half an hour.
- Restarted Opera and hey presto, the forum shows. Might have restarted Opera earlier, don’t remember, but maybe the cache now timed out and did a proper refresh.
Funnily I woke up this morning and Opera did yet another upgrade. Maybe this hard cache upgrade created a lot of problems elsewhere too 
.
Additional note: 3) is of course the most interesting part for us. Maybe it can still be found from the logs (can dig if needed). If Discourse crashes, does it try to relaunch using some timer? Or was the whole site down until I logged in and relaunched the app?