@dhdurgee: the issue is not how *we* can ID them, but how does the *framework* does it, and if we can catch that and link that back to a physical file. Supposedly it does, and hopefully in a unique-ish manner, since it's in the DB, but then again, we're not relying on the DB keys here, so all may not be rosy.
In any case, that would only ever be for manual cover processing, PDF is a whole other kettle of fish compared to Mobi, so pulling & processing covers automatically, even if there were any (which, I'm fairly sure there aren't, which would leave processing the first page) is out of the question
.