The "oddness" I sometimes see, I think is not actually the fault of hathitrust, but restrictions placed on full download might be put in place by the institution that sponsored the digitizing. I have often seen multiple offerings of the same book, with some being restricted to a "partnership" download, and one being freely available to anyone. Rather odd.
I have also noted that Google pdf scans posted to archive.org are seldom OCR'd --- ??? pirate versions of google-scans ???
(I should say, they have no text layer included in the pdf. There will usually be a separate download of the OCR "full text".)