Brendan Nyhan makes sense:
[T]he notion that periodic individual elections can create permanent changes in partisan alignments is not well-supported by the historical record, which looks far more messy and chaotic. As the Yale political scientist David Mayhew and others have shown, even the supposedly classic example of a realigning election -- the 1896 election of William McKinley (Rove's favorite) -- cannot withstand empirical scrutiny. It's hard to imagine why 2008 would have been any different.
Given this context, the fragility of Obama's majority should not be surprising. Indeed, it should be reassuring from a democratic perspective. The last two elections swept many Democrats into office who were relatively poor fits to their district. In some of those districts, the replacement of those Democrats with Republicans will create a better alignment between the views of voters and their representatives. In other districts, the person elected may not match the voters well and will face contested races in future challenge. In general, there is little reason to think that the pro-GOP shift will be any more permanent than Democrats' gains in 2006 and 2008.