At longer exposures it is more important to do a dark frame subtract, and the question is how repeatable the thermal noise profile is. A dark frame subtract is pretty effective if (for example) you have a brighter corner, but is not good with individual pixels that randomly go hot as noise.
I don't know about D800 itself, but it does bring up the question whether very high resolution (meaning more pixels for a given visible portion) in and of itself with similar noise profiles makes the dark frame more or less effective. Trying to convince myself one way or the other as a thought experiment and not sure. Any astronomers around?
PS. This darm substract is what the long NR setting causes, it takes an exposure while the shutter stays closed, then subtracts it from the exposure with the shutter open. Sort of (it doesn't take exactly the same length exposure, which is the more traditional approach).