Just set a cron job up to scrape and md5 all of the documents every 20m or so; if the MD5s match, discard. If they do not, save. Then you can build a timeline of releases, redactions, edits, and fuckery.