Hello,

I know there is not much you can do about this, but I want to raise awareness. For the past approximately 2 weeks, the CI that runs on Pull Requests is pretty much unusable for packages with builds that take 1+ hour. As Koji and src.fp.o occasionally give 50x errors (see e.g. https://pagure.io/fedora-infrastructure/issue/12814) and the CI jobs heavily rely on koji scratchbuilds from dist git, every time the CI runner gets 50x it seems to abort (instead of retrying to fetch the build status, retrying to download etc.). When a job waits for a ~5 hours build to finish with polling or a blocking koji command, it's almost certain it gets a 50x during that time at least once -- effectively making the CI job very fragile.

It has been a very unfortunate experience for me, where we ned to [citest]/recheck builds multiple times in a row and it takes 2 to 3 days to successfully verify a change over CI.

Packit CI is a tad better with this, it seems, but not completely resistant.

Some examples include https://src.fedoraproject.org/rpms/python3.15/pull-request/1 to 8 or https://src.fedoraproject.org/rpms/python3.14/pull-request/112 to 122.