Corrupted Slave Data?

[Tim Daneliuk]

Running bind 9.16.15 on FreeBSD 11.4-STABLE.

Master is out on a cloud server at Digital Ocean.  Slave is on-premise.
All on-prem LANs point to the slave instance.

Running split horizon to keep nosey parkers out of our local DNS assignments.

Recently - and for no obvious reason - the on-prem instance stops resolving
properly.  The fix is to stop it, clear out the slave files, and restart.
Then it works for a few days and repeats its misbehavior.

The logs show nothing remarkable, at least at first look.

Is there a known slave file corruption problem?

Could someone kindly suggest things we could look into otherwise?

Many Thanks ...