My inspection method is to skim through and open the files. The bulk of the data I'm backing up (images for the most part) are all very similar in file size, so that makes an easy way to spot critical backup errors. If all the files are 5kb rather than twenty some MB, then I know something obviously went wrong.
I also generate a hash manifest of the files in the archive. This file is easy to confirm as still valid as I can check that the xml tree is still intact, and then compare the redundant metadata. Would need to dig out some notes to confirm details, but the short of it is that each file gets an entry in the manifest as part of the automated archiving system:
- File name/location
- Hash of the file's data, excluding file metadata of things like "date last accessed" that can change, which is a short code of characters generated based on the data.
- Error detection strings of file name and hash value. Short mathematical code that can be compared.
The odds of a file error keeping the XML intact are terribly low in and of themselves. The odds of a file error also changing anything in the above values in a way that still line up and not raise a flag are insanely low.
However, the deep inspection of regenerating the hash values for everything in the archive does take time: Unzip to temp copy, run data tests, compare to manifest, does take time. Luckily it still runs through in less than 10 hours, so it mostly means leaving a computer to run over night a dozen times a year.
However, just loading up the data and looking through the photos is probably 'good enough' for the majority of people. Be as detailed as you feel you need to be for your data, but just make sure it is actually there. Far far too often I see stories about people who only ever reach for their backup files when there is a problem. And when you have a problem and you need your backups is not the time to discover that there was already a problem there as well.
Canon EOS 7D | EF 28 f/1.8 | EF 85 f/1.8 | EF 70-200 f/4L | EF-S 17-55 | Sigma 150-500
Flickr: Real-Luckless