Example: https://gitlab.com/redhat/rhel/rpms/389-ds-base/-/merge_requests/185

The original pipeline in the MR from 6 days ago (as of 10th March) was successful, and there were other successful MR/merge train runs of check-tickets for that stream branch from around the same time.

Today, the merge train was failing: https://gitlab.com/redhat/rhel/rpms/389-ds-base/-/jobs/13391362202.

Retrying the previously successful job in the MR was successful, but running a fresh pipeline failed. This implies there was an update in the image in the last 6 days that introduced a regression and the smoking gun is centpkg, as we rebuilt the image to pick up the latest version to fix some other problem.