This is a bit of a closed thread, but I just wanted to mention that the archive does actually archive non-HTML pages. Not sure if it's accidental or whatever, but this philosophy will probably cover all bases: If it looked like something just downloaded from the web archive and it worked, back it up in at least 16 places.
As an aside, I've been noticing that some files seem to be truncated at $SIZE - 1 bytes, so for example wgetting a 142368 byte file will cause wget to notify me of the connection being closed at byte 142367. :-/ [?]
-dav7