One trouble with testing restores is that my likely avenue in the event of a real catastrophic failure that required a real restore would be to have them send me the data on a hard disk, to the tune of about $200. Not something I want to try just to make sure it works. I could try a download restore, but this would entail downloading a couple hundred GB of data, and wouldn't really exercise the same path I'd use in a real scenario. Any thoughts on the right way to approach it? Doing occasional small restores (usually using Backblaze as a sort of really slow Dropbox, restoring to a different computer when I need a file I forgot to share) at least gives me some peace of mind that the system works at all, but you're right that it's not quite rigorous.
For servers, I've heard good things about Tarsnap, but never used it myself.
No, that's a tough one, especially with a back up the size you're doing.
Everytime you do a restore that's just confirmation that a particular backup was/was not sound.
Wouldn't it be neat if the back up program had functionality that allowed you to restore a random sample of your files ( and maybe then compare these to the originals ) and give you a likelihood that your entire backup is sound?
A sort of sampling quality control process.
For servers, I've heard good things about Tarsnap, but never used it myself.