On CI, the Azure pipeline (for instance) produces a zipball of logs. This contains some of the systemd journal as well as a couple things in /var/log. However, it doesn't include the krb5kdc logs. This makes CI failures hard to diagnose for me.
It seems like this might be a "sometimes" problem, possibly when the KDC doesn't start correctly. (I just got a tarball that did have a krb5kdc.log in it this time.)
Metadata Update from @rharwood: - Issue close_status updated to: invalid - Issue status updated to: Closed (was: Open)