Make withdrawals foolproof
Recently, a student accidentally withdrew a submission when he only wanted to withdraw a resubmission because he forgot to reply to referees while resubmitting. It would be nice if such mistakes were harder to make and easier to repair.
For a start, withdrawing a submission should just mark it as withdrawn, without removing any publicly available information. The author should then have the options to ask that the webpage be removed, or to resubmit the withdrawn submission, or just to leave it as it is.
Moreover, even if the submission is removed from public view, it should still be accessible to authors, referees, editors.