My initial guess is it's the classic problem of collusion. Everyone agrees to do something, in this case to vote "Yes," but everyone has a political incentive to "cheat" and vote "No." If enough people cheat, then the legislation doesn't pass.
(Wikipedia: Cheating: There is considerable incentive to cheat on collusion agreements; although lowering prices might trigger price wars, in the short term the defecting firm may gain considerably.)
Ah, Congressmen behaving rationally and leading to a sub-optimal outcome. Perfect in-class example.
UPDATE: Looks like I might be right. Over on The Economist's blog:
"The word on the politics of the deal is this—Congressional leaders had come to an understanding whereby "safe" members of both parties (those not facing a serious challenge in November) would vote for the bailout and endangered members of both parties would be able to vote against if they chose. For the most part the Democratic breakdown proceeded as anticipated. The Republican caucus, on the other hand, does not seem to have delivered the expected "yea" votes. So, strictly speaking, GOP defections were the proximate cause of the defeat. Whether it will appear that way to voters, amid so complex a legislative scenario, is unclear."