It is CW that an incumbent under 50 percent is in danger. But just because it's CW doesn't mean it's true. So let's look at the actual numbers from 2008 and see if the rule holds up. I'll stick with Senate races since they have more polling available.
In the 2008 cycle, the following Senate races featured incumbents with at least one reputable independent poll showing them under 50 percent against their challenger: Alaska, Georgia, Kentucky, Louisiana, Minnesota, Mississippi, New Hampshire, New Jersey, North Carolina, Oregon, and Texas.
Of those 11, the challengers won five: Alaska, Minnesota, New Hampshire, North Carolina, and Oregon. That's a 45 percent success rate.
All other incumbents running for reelection, those who didn't suffer poll results under 50 percent, won handily. That's a zero percent success rate for their challengers.
In the 2006 cycle, the following Senate races featured incumbents with at least one reputable independent poll showing them under 50 percent against their challenger: Arizona, Connecticut, Florida, Michigan, Missouri, Montana, New Jersey, Ohio, Pennsylvania, Rhode Island, Virginia, and Washington.
Of those 12, the challengers won six: Missouri, Montana, Ohio, Pennsylvania, Rhode Island, and Virginia. That's a 50-percent success rate.
So over the last two election cycles, no Senate incumbent polling exclusively over 50 percent has lost his or her race. And of those polling at least once under 50 percent in a reputable independent poll, 47.5 percent of their challengers have been able to knock them off.
It may be a small sample size, and perhaps this is a topic that needs Nate Silver's statistical wizardry for any solid conclusions. But from my cursory look, it definitely appears that in this case, the CW is correct. Incumbents over 50 percent in polling are safe, and those under 50 percent are a coin toss away from defeat.