Nate Silver at FiveThirtyEight looks at National Popular Vote, and some Republican vetoes of it, and notes:
People, obviously, are going to remember 2000 for a very long time, in which Al Gore was screwed by the Electoral College. But throughout most of 2008, our simulations showed that the Democrats, not the Republicans, had a structural advantage in the Electoral College, something which was also apparent in 2004 when John Kerry nearly won the Electoral College in spite of trailing in the national popular vote by 2.5 points.
Indeed--and I don't have evidence at hand or time to dig at the moment--in the runup to Election Day 2000, there was more discussion of Bush taking the popular vote with a Gore electoral vote win, rather than the way it really happened. A split result is most likely in the last election of a census (like 2000, or 1888) because the discrepancy between census apportionment and actual vote is bigger. So the way it went wrong in 2000 was counter-intuitive: the popular vote winner took the states that were about to lose electors. Bush gained five or six electoral votes in 2004 from reapportionment
What would it take for there to be a real chance of abolishing (or end-arounding, as the Compact seeks to do) the Electoral College? I think it would take two elections in relatively rapid succession in which there's a popular:electoral split, particularly if these two elections are won by candidates of opposite parties. The memories of 2000 should linger for a few more cycles, and so if there's another such occurrence before, say, 2020 or 2024, things could get very interesting.
A 50 state strategy (OK, maybe a 40 state strategy for Obama in `08) made an Electoral College mistake less likely. To win while losing the popular vote, you have to win the squakers and lose the blowouts, and when you work more states the blowouts aren't as bad.