Oh, another person also mentioned using ECC memory, which may not be awfully important until it suddenly is.

Right now i have incremental backups over the network with BackupPC: https://backuppc.github.io/backuppc/

That software solution sort of dances around the issue, because if any files were to become corrupted, i could just go back in time to their older versions, though that approach isn't necessarily good either - since that means having to store the original "full" backups for a long time and also takes up more storage, which i do for the essential data, but not the stuff that i'm willing to lose. And, since the data stored can be arbitrary and therefore checking for corruption would have to be done manually, there are serious drawbacks to that, in regards to even knowing when things have gone wrong.

Of course, there's also the possibility to use additional sources of redundancy, like versions within Nextcloud should the actual file contents of a particular version become corrupt, however for certain scenarios file systems like you've described indeed do become the way to go. Maybe not for every homelab out there, though.

yea, error correcting memory is a second step towards data integrity. but it is of little use if your at-rest storage has no means to detect errors (like normal filesystem or raid1).