The typical process that happens in graduate application selection is that you subdivide applicants into three groups:
- Accepted applications: These are your first choice of candidates, and you offer them a position. This list need not be ranked.
- Waitlist: These are the applications you think are good enough to accept, but not so good that you want to make them an offer right away. You rank this list, and go top to bottom as people who have previously been offered a position decline.
- Reject: These are the applications you think aren't good enough to succeed in the program. There is no need to rank these candidates because you don't anticipate ever tapping into this pool.
The point I'm trying to make is that the rejected applications are candidates you don't think are good enough to succeed in the program. You're probably going to ask what happens if everyone declines and you're out of names on your waitlist? Well, then you just don't make any more offers. That's because you've previously decided that the ones on the "Reject" list aren't good enough to succeed.
The only thing that's worse than not being able to fill all graduate student slots is to fill these slots with students who you believe will not succeed in the program. That's a sunk cost: you're paying for people who aren't good enough to do the job they're paid for.
So, under the assumption that all universities work on such a system (which I think is, in essence, the case), there is no reason ever to turn a rejection into an acceptance.
(Now, has it ever happened? Almost certainly, given just the sheer number of cases. Does it happen on a regular basis? No.)