Dienekes comments on the paper which showed genetic substructure among a set of Sardinian villages:
The take-home lesson is that wherever gene flow is impeded, no matter how geographically close, population differentiation can be recovered with dense autosomal genotype data. Really fine-scale ancestry analysis is now possible; I suspect that a combination of geography, religion, social class, language, and ethnic identification will be found to be predictive of a person's broad genetic makeup and vice versa. But, to discover these correlations, a large-scale collection of genotypic data is required.
The potential exploration of population structure mapping onto non-geographic variables is definitely going to be interesting.