This test has probably been broken for a very long time, but it was missed due to UPGRADE-ASDF returning T all the time. This meant that the first LOAD-OP was performed in a new session, *not* the session created at the top of the file. Once UPGRADE-ASDF was fixed, then the done status of the LOAD-OP was persisted, resulting in the next LOAD-OP being a no op. However, this breakage wasn't immediately apparent because FILE-WRITE-DATE was compared with #'>= and the PLAN was not asserted to actually contain the operations we wanted to see. So on a fast computer (unless you got really unlucky) you'd never see the failed test.