It now seems pretty clear that virtually all of the late polling on the Democratic side was wrong… very wrong. The last Rasmussen Report had Obama +7 over Clinton. CBS had him +7. USA Today had Obama +13 and CNN +10.
With more than 25% of New Hampshire now reporting Obama is -5. 80% of precincts still need to report. Nashua still needs to report. But this gap really is extraordinary. Chances are nil that Obama is going to win overwhelmingly.
Why?
It is a return to the race-gap polling problems of the 1980s and 1990s:

This phenomenon was first noticed in the 1982 race for governor of California, where Los Angeles Mayor Tom Bradley, a black Democrat, narrowly lost to Republican George Deukmejian, despite polls showing him with a lead ranging from 9 to 22 points. The next year, African American Democrat Harold Washington barely won his race for mayor of Chicago against Republican Bernard Epton. Pre-election polls taken within the last two weeks of the campaign showed Washington with a 14-point lead.

The problem was prominent in the New York City mayoral race in 1989. David Dinkens, an African American candidate beat Republican Rudy Giuliani by only 2 points, despite leading by as much as 18 points in polls a week before the election.
Tonight, despite all the talk of how little race matters in this campaign, it is clear that race is still a big deal in bi-racial campaigns. And it has showed up for the first time, in a measurable way, in the 2008 presidential race.