I'm not sure if koji is to blame here, I think it just sets the timeout in mock? But I could be misremembering...

In any case I agree it would be good to get this working again.

Timeout and restart happens in different cases. Restart means, that something bigger is going on. Build hadn't reached set timeout, but faild in som non-standard way (something what was not noticed by rpmbuild/mock) - like failing builder, kernel issues, etc. Are we able to get kojid/journal logs from around the time of restart? Anything suspicious there?

a recent example of many times restarted buildArch task = https://koji.fedoraproject.org/koji/taskinfo?taskID=69040544

So, this one is various things:

I was reinstalling builders yesterday, so I freed it several times to move it off a builder I was re-installing.

I can't tell much on the later ones. No OOms on that builder, no kernel issues that I can see. No ideication of why it restarted. ;(

I was briefly watching this one and when the gcc test-suite was run, it got restarted around noon CET. And seems it got restarted a short while ago too ...

a new candidate to investigate = https://koji.fedoraproject.org/koji/taskinfo?taskID=69702088
(and there are more right now)

This is a list of tasks not progressing I have cancelled today, the oldest were "running" for ~2 weeks
69701974 70093944 70106858 70228793 70247868 70347362 70473614 70489935 70573749 70524504 70683005 70706039 70612859 70615113 70618949

This is still happening. ;)

https://koji.fedoraproject.org/koji/taskinfo?taskID=108910932 is a recent example.