All that is true. How does it relate to the possibility or not of large differences between human populations? Interbreeding will reduce variation extremely quickly but we know reproductive isolation was very high between continents for a very long time. Even individual villages in India have jatis that have been living next to each other for centuries that haven’t interbred for over 1,000 years.

Genes run in families, not villages. How are the traits spreading across the entire population in that 1000 years enough to make the "large difference" instead of being averaged out again? Existing cases of that are founder effects or mating with someone truly different like a Denisovan[0].

More importantly this knowledge would be useless if you found it, because the only reason to know things is to make predictions, and this doesn't let you predict anything about the next generation. We're in modernity now and they can just move somewhere else.

[0] which funny enough isn't considered a subspecies and doesn't have a Latin name because the scientist who discovered them thinks Latin names are pretentious.

> How are the traits spreading across the entire population in that 1000 years enough to make the "large difference" instead of being averaged out again?

Mating isn’t random. People are more likely to mate with those who live near them. So Southern Chinese are more like each other than Northern Chinese. Geography matters. Distance matters. If there’s reproductive isolation difference can emerge and persist. If there isn’t it won’t.

Geography isn’t the only way populations stay distinct. Religion, language, means of sustenance can do it too but interbreeding can reduce difference very, very quickly as you note.