Edit: postponed by one week most likely.

I might have identified the cause of the occasional stalling that you likely noticed when using this Lemmy instance.

It seems that during one of the previous migrations and restoring the database from a backup the noCoW flag on the database files was reset and this causes slowdowns on a btrfs file-system specifically for databases.

To fix this we need to recreate all the database files with this flag set, which requires a somewhat longer down-time, maybe 3 hours or so.

While doing so we can also do a database vaccuum and upgrade to Postgres 17, so even if this turns out not to be the real cause of the stalling we would still get some benefit from it.