And I don't think the patent office can refuse to do a reexamination on a patent.
I want to clear up a couple points of confusion that have sprouted up in this thread. First, BC2009 - great post and you're 100% spot on. This is a non-event. But I want to address this point. The Patent Office CAN deny a reexam if there is no Substantial New Question of Patentability (often shortened to SNQ). In the past, denials rarely happened, but the Patent Office has been pushing back in the last couple years as examination has gotten more thorough (a lot of patents examined in the mid-to-late ninties and early 2000s were let through too easily, so there are more SNQs as prior art comes to light; this hasn't happened as much lately, so there's fewer reexams granted).
As for obviousness vs. prior art as a means for rejecting, they are not mutually exclusive. There are primarily four hurdles a patent application has to overcome to be issued. The numbers below are for the sections of 35 USC they are related to
101 - utility: is it actually possible and useful. You can't patent a perpetual motion machine because it isn't possible. The useful part is the lowest of the low bars.
102 - novelty: is there one piece of prior art out there that does exactly what is claimed. If there is one piece of prior art, the patent claims aren't novel.
103 - obviousness: are there multiple pieces of prior art, that when combined, have all the claimed limitations? If there's one piece of art, the examiner will use 102. If he finds two pieces of prior art that each don't have all the claimed limitations, he will say "a person having ordinary skill in the art will take A from 1 and B from 2 and put together. Doing so would be obvious for that person." So an obviousness determination is BASED on prior art and what a person having ordinary skill in the art would know.
112 - written description/enablement: this actually has a lot of subparts I won't go into, but basically the applicant has to tell you how they did it. There is a lot of wiggle room as to how much is enabling and there's a weird line between where you in effect stop describing the details because it would be obvious how to do it (I can say "and then the blah blah queries a database to retrieve X" and I don't have to describe the actual query because that's obvious, but then isn't the step of querying as part of the combination obvious??).
Under the old reexam rules, you could only submit a request for reexam IF you had a new 102 or 103 rejection. Under the new rules, I believe - I need to look this up - you can submit it under any grounds, but they are vastly limiting the reexam process under the new system in favor of a "post grant review" process.
*whew* sorry for the wall of text. Hopefully that clears up some stuff. Yes, I am a patent attorney.