If this comes up again, ask the user if they disabled copy-on-write for their database. There is really bad advice out there about btrfs and databases. Users commonly (and incorrectly) do
Code:
chattr +C /path/to/database
This is practically the mantra. But this disables CoW (and all the data integrity features btrfs employs). The damage is silent until a triggering event like replication, balance, or defrag.
I have been using Calibre on a multi-disk btrfs system for years, and the only problems have been self-made. Aside from the time I believed it would be wise to convert my Kobo's filesystem to btrfs, the worst experiences came from disabling CoW for things like databases and VMs.