I found the line about Tigerbeetle's model assuming entire disk sector errors but not bit/byte errors rather interesting - as someone who has created error correcting codes, this seems out of line with my understanding. The only situation I can see it works is where the disk or driver encodes and decodes the sectors... and (on any disk/driver I would care to store an important transactional database) would be reporting tonnes of (possibly corrected) faults before Tigerbeetle was even aware.

Or possibly my mental model of how physical disks and the driver stack behave these days is outdated.