[I]t looks like most Indian jatis have been genetically endogamous for ~2,000 years, and, varna groups exhibit some consistent genetic differences.
The level of this endogamy at the jati level is extreme. I personally wonder if some of that is due to non-endogamous individuals being jettisoned from their caste, as it is stunningly hard to maintain that level of endogamy (on the order of 99.9%+ compliance in each generation) for two thousand years in people who live cheek by jowl in the same cities, villages and regions, and have overlapping appearance phenotypes, without such a purifying mechanism.
There is lots of structure and diversity in the overall population of Pakistan-India, more or less on a northwest-southern cline, as well as by varna. At the varna level, this is mostly due to differing degrees to steppe ancestry, although there is a deeper level of Iranian Neolithic farmer v. South Asian hunter-gatherer cline that runs in parallel along very similar geographic clines in South Asia.
Bangladeshi people have essentially the same amount of South Asian ancestry which isn't very diverse, and considerable variation in Tibeto-Burman ancestry. This probably has something to do with frontier founder effects and the way the frontier destabilized traditional social organization.
Pakistani people, genetically, have a genetic mix very similar to that of people from India, despite the fact that as Muslims, they do not give religious credence to the caste structure of the Hindu religion.