As the HN link you posted demonstrates, query strings quite often determine the content. You can't strip them out without introducing reams of false positives.
What about HN also storing the page title ? Upon submission, strip after "?" to check for dupe, in case of positive, check page titles to confirm.
Or just check, upon first submission, that the query string is irrelevant (fetch pages with and without), and strip it out if it is.
A bit slower, on submission, but that wouldn't be bad ;-)