Just a reminder, folks.
Corrupted backups are just that.
In other words, when the backups are run, there is a signal put out that a successful backup was done. If THAT part of the safety net breaks, it's nearly impossible to catch.To catch it, you have to restore a backup. And, at that point, it's too late. Having even 10 copies of each backup doesn't matter, if the software making the backups corrupts every one of them in the process of making them and saving to disk. It's a problem as old as magnetic media itself.
After all, look what happened to Stratics in Spring 2008.
To give a real-world analogy...
When the Hubble Space Telescope was being built, they had several levels of redundancy. However, all those redundant checks were being done by machines that were all kept in calibration by a single device - one that was (somehow) OUT of calibration, when that was supposed to have not been that possible. As a result, the constantly "backed up" error correctors were CREATING errors, while everything seemed to be perfectly in place.