> Otherwise, the database knows from the ID of the journal not to reapply an already applied rollback journal.

But it's not "already applied", that's the whole point. The transaction was committed, not rolled back, so the changes in transaction were persisted to disk and the journal was just thrown away. If it magically reappears again, how is SQLite supposed to know that it needs to be discarded again rather than applied to revert the change?