According to conventional wisdom, delegates from FL and MI will be seated if the votes of those delegates can't change the outcome of the race. But what will happen if Obama is narrowly ahead, such that, if FL and MI were counted, Clinton would win the nomination? How could the DNC preserve the outcome determined by its own rules without alienating voters from two huge swing states?

There is an excellent op-ed in today's NYT (http://www.nytimes.com/2008/02/15/opinion/15mann.html), which argues that historically, superdelegates have acted as a unifying force, reinforcing rather than rejecting the winner of close popular votes.

In 1984, the superdelegates stepped in to provide a majority for Walter
Mondale — who had a huge edge in pledged delegates over Gary Hart but
not enough to win the nomination — avoiding a potentially bitter and
divisive convention that would have fractured the party.

The piece goes on to suggest that superdelegates can solve the FL/MI dilemma:

In this case, the nomination could come down to a difficult and complex credentials
battle over whether to seat delegates from Michigan and Florida. To
have a nomination settled in this way is a bit like having an election
settled by a 5-4 vote of the Supreme Court. Averting this kind of
disaster is just what superdelegates are supposed to do.

So how would this work? Remember, the DNC needs to preserve the outcome of the race without MI and FL yet somehow include their delegates in the race. Howard Dean could find superdelegates who don't have strong preferences for either candidate but are very concerned about alienating FL & MI voters to pledge for Obama in order to balance out the FL & MI Clinton votes.

How many would he superdelegates would be needed? First, let's do a back-of-the-envelope count of FL & MI delegates. According to CNN, if all their delegates were counted, FL would have 210, including superdelegates. FL voters went 50% for Clinton and 33% Obama, about 3 to 2. For simplicity, let's assume that the Edwards votes and superdelegates are a wash and split the total delegates according to the 3-2 ratio: 126 to 84, a difference of 42. (If it were to actually happen, Obama would almost certainly do better than that among superdelegates and Edwards supporters, picking up a few more delegates).

It's a little unclear how to handle MI, since Obama wasn't on the ballot. But to be conservative, let's give Clinton delegates for her 55% vote and ignore the 40% uncommitted votes. If MI's delegates were counted there would be 156, so that's 86 more delegates for Clinton. (I know, it's a little absurd. Maybe the DNC would find a way to give Obama some MI delegates.)

In any case, even with such conservative estimates, Clinton would pick up a maximum of 128 votes. Let's suppose that Obama has a 50 delegate lead over Clinton without FL & MI. In that case Dean would need to come up with 78 delegates for Obama in order to preserve the outcome. Since roughly half of the uncommitted superdelegates would likely have voted for Obama anyway, you need to double the number to 156 or else find 78 Clinton-leaning superdelegates who are willing to vote for Obama in return for counting FL and MI. If Obama has a larger delegate lead or gets more delegates from FL & MI than my conservative estimate, then that number would be even smaller.

Crisis averted. FL and MI voters get their votes counted, but the outcome is the same as it would have been had they not been counted.