The ones I can recall right now:
Rotating log files without restarting the service. File is deleted while still being open by Apache or Tomcat, "du" shows 10% /var/log filesystem usage, while "df" says the usage is 100%.
Mounting NFS for temporary usage and forgetting to unmount. The NFS server gets rebooted at some point, so the NFS mount hangs. Admin scripts which depend on the storage (like checking free space) begin to hang as well, so their functions like cleaning directories and log rotation gets paused too. And if someone forcefully unmounts the bad NFS mount during business hours, show begins: released from the hanging state, log rotation scripts start to do their job at 1pm instead of 1am
Security improvements is a special topic. For example, specifying a group name with allowed SSH login in sshd_conf. Non-members of this group can't log in via SSH anymore, and this cause isn't easy to find.
Using immutable flag and ACLs with files may be impressive for those who don't use them - and most ones don't.
Backup files in the drop-in config directories, like /etc/sudoers.d ! Sysadmins forget than both old and new versions are active, so they overlap and create unexpected effects.
