I have a backup slave server, where I take regular snapshots of the database via LVM.

By default, with relay_log_purge enabled, the relay log files get deleted as soon as the server has processed them. I thought it might be handy to hang onto them, in case they were useful if the need to restore from these snapshots ever arose, so I disabled relay_log_purge.

I was expecting the relay log files to be automatically deleted once relay_log_space_limit was hit, but instead once that limit was hit, new relay logs stopped being written and replication ground to an ugly halt.

It seems to me that the behavior should be to cycle out old relay logs when relay_log_space_limit is hit, or if that is unset, when the partition fills up. That would allow me to keep as much relay log "as possible" while also not having to manually (is that even possible/recommended?) delete old log files.