When you see checksum errors into the millions, it’s a sign of poor cable connections or an LBA issue.
Possibly related, you had been writing, reading, and checksumming data while you had bad/failing RAM installed, possibly as far back as November. This means that if “corrupted” data was written to disk (because of bad RAM), it will not show up as corruption with a ZFS scrub. The checksum will confirm that the block written to disk has the expected checksum.
An app has no effect on the data and checksums that were already written via ZFS.