Non-ECC RAM is fine for a Celeron NAS.

Probably not. ZFS checksums catch bit flips in RAM before they hit the disk, so silent corruption from non-ECC memory is more of a theoretical risk than a practical one for a home NAS.

The whole “ECC or bust” thing comes from enterprise paranoia. In theory, a cosmic ray or voltage fluctuation can flip a bit in your RAM, causing ZFS to write corrupted data that its own checksum then validates as correct (since the checksum was computed after the flip). But in practice, the error rate for modern DDR4/5 RAM is something like one bit flip every few years per gigabyte of RAM, even without ECC. On a Celeron-based NAS with 8-16GB of RAM, you’d be unlucky to ever hit it.

ZFS already does far more than typical filesystems to protect your data. It scrubs, checksums on read, and self-heals on mirrors/RAID. If you’re really paranoid, run a weekly scrub and you’ll catch almost anything. The bigger risk to your data is a bad disk or a power outage, neither of which ECC helps with.

Save the money you’d spend on ECC RAM and an enterprise mobo, and put it toward a UPS or an extra backup drive. That’ll protect your data more than ECC ever will on a Celeron.