Gotcha. Prunes are scheduled to run no more than once every 7 days. (We’re gonna improve the logic in the future so that for larger repos, it happens less often, but also subject to desired retention periods.)
Prune runs only from the client, since we don’t have the repository password.
Prune does not run on a repository if there are any started-and-unfinished backups to that destination. This is the tricky part that a lot of simpler scripts don’t consider, potentially causing data loss if an initial backup takes a while: See Quick question about running prune.
You can read more here about how retention periods are enforced: https://relicabackup.com/docs/retention-periods