I think the challenge to it is that DHT structures right now are not themselves commutative. We need a CRDT foundation for DHTs, themselves!

And one that has an asymptotic algorithmic complexity bounded near O(N^2) or less to handle the massive workloads these things can see.