I'm seeing some trends here:
1) Failure to do "restore testing". When you have RAID, preferably run consistency checks (if supported), but _always_ test with a "failed" (removed) drive.
2) A generalized lack of integrity testing.
Both of these can be accomplished by the following (tedious) procedure:
0) Verify your backups can restore _first_; on a RAID, this procedure is designed to show you problems in your RAID consistency if they exist.
1) Run Jacksum (with the -m option) or some other subdirectory searching message digest generator/checker on all your data. Store the result file somewhere else.
2) "Fail" a drive (or restore your backup media to another previously blank device, or mount your directly readable backup media)
3) Check all those message digests with Jacksum (-l -c) or the other equivalent.
4) Ideal for RAID: put a blank drive in and let the RAID rebuild, then repeat step 3).
For those backing up to CD or DVD, there are a few pieces of software (DVDisaster, ICE ECC, etc.) that allow the use of some of the DVD's capacity to allow for error detection and recovery due to damage or degradation of the media.