Joe Romm has a pair of very good (and very disquieting) posts revisiting the question of permafrost feedbacks. The basic idea is this: As the planet heats up and the tundra regions start thawing out, all that methane and soil carbon that's stored beneath the permafrost is going to start bubbling out into the air, which in turn will warm the planet even further. A potentially nasty cycle.
It's notable that the 2007 IPCC report on climate change didn't include the feedback effect of permafrost melt in its already dire projections (there were too many uncertainties). When skeptics start talking about how climate models aren't perfect and don't take into account this or that… well, they're not entirely wrong, but there's very good reason to think that, if anything, the consensus underestimates just how bad climate change could get, rather than overstating it.
Anyway, Romm argues that this means we're going to have to try to stabilize atmospheric concentrations of CO2 below 450 parts per million, because if we shoot for even a slightly higher target—say, 550 ppm—it could prove difficult or impossible to stop chain reactions like the tundra feedback from kicking in, at which point runaway warming won't be out of the question. See this post by David Archer on the subject for a lot more detail (scroll down to the "permafrost" section). Plus, of course, this is hardly the only feedback effect to fret about...
P.S. For those curious, that weird-looking photo is of methane bubbles trapped beneath a frozen lake...
--Bradford Plumer