Did the backup jobs themselves become corrupt?
I've reduced our two proxies down to 4 threads each, this seems to have fixed the issue we were previously seeing. Apparently each CIFS container on the NAS can only handle 16 threads anyway. Previously I had 2 x 12 threads firing at it, so this may have made things worse. I've done a couple of test restores from the data, it seems OK so far.
What environment were you testing it in? 1Gb/10Gb? How many threads/proxies?
Thanks for your feedback. Interesting to hear someone else has been down this road