Genetic drift occurs in traits where there is no selection pressure. Skin color, for instance, is selected relatively weakly, so it can be locally adaptive after a long, long time. But I think it's pretty clear there's one human trait which has been under the most intense selection pressure in the modern era (~10,000 years at least, but obviously going back much further): intelligence.

I don't believe it is possible for substantial sub-population variation to exist in traits that are heavily selected for. Anything related to sociability, for instance, is right out. I feel that even measuring slight differences wouldn't be enough to overcome this prior (although, of course, when you properly disentangle things, no such difference is found in e.g. the children of WW2 GIs of different races, so measuring doesn't find these differences except in the most slap-dash sense of 'measure').

Perhaps this is a misunderstanding of Darwin. I know his finches found multiple, concurrent, stable niches. But that's because the optimal strategy (for finches as a whole) was mixed, or variant across space. It makes sense to have a population split into two more specialized populations. Gradient descent can diverge when there are optimal mixed strategies. But intelligence really, really isn't that. Socialization isn't that. If someone manipulates you and you are a sucker, you lose. If you think of good ways to cooperate, you win. Perhaps a more thorough look at resources would change my mind, but this is my current understanding.

[Also, I mean, wowzers are the people saying the opposite losers. Unrelated, but it is the case. Some are not idiots – Charles Murray is not an idiot, I don't think – but it's extremely low status. I think that while it's tempting to make a 'status argument' – or, just make fun of them – it's also worth engaging on the object level, and saying this isn't plausible.]