Even without using ZFS (I prefer XFS as significantly faster) all the files that I store have content hash values in extended attributes, for data integrity verification (and also for data deduplication).

Whenever I write that drive, after a power cycle (to be sure that the files are read from disks and not from some cache) I run a script that checks the integrity of the files, to be sure that I can remove them from elsewhere without risking data loss.

With SMART-enabled drives, I usually do that only in the rare cases when a drive reports corrected errors, because I have seen cases when a drive miscorrected some errors, resulting in corrupted files. With a HDD without SMART, when the drives finds errors, but it believes to have corrected them successfully, there is no external sign that something could have gone wrong.