You are not mistaken. Reopening a set after the due date shouldn't change the problems unless you manually change seeds. If the problems really did change I would expect something else to have occured.
If you edit a problem and add a new call to random somewhere, that might change some of the other values in that problem, but not other problems in the set.