@nprofile1q... neither.
For recovery after failure of individual sectors, ext2/3/4 is king, precisely because of the things that xfs proponents keep insisting are "design flaws", like static allocation of inodes at filesystem creation.
Random bit flips aren't an issue, because there is ECC information on the raw media, so you will see defects in the form of entire sectors failing to read. If you see bit flips, it's just more likely you have bad RAM on your host.
If you can schedule a regular readout (just read the entire block device), that gives drive firmware an opportunity to spot ECC-correctable errors and silently fix them. Whether it actually does, depends on drive firmware.
Other than that, the real advice is to not do long-term storage on shitty SSDs, especially as their failure mode is to suddenly become read-only, so after they failed, you can't even wipe them before throwing them away.
Every friend group should have one person with lots of storage, either RAID6 with harddisks, or a tape robot.