I've been closely following David Barbour's progress since early 2011, occasionally making my own attempts at implementing the basic evaluation model. As far as I'm concerned, RDP is far and away the best upcoming approach to making secure and physically viable programs.
I'm open-minded to improvements upon RDP, but RDP hasn't matured enough to say for sure where it ends and the improvements begin. :-p So I've been focusing my language design efforts in places that would mesh well with however RDP turns out, due to orthogonality.
Namely, I don't strive to model world state, multi-user real-time interaction, or even general computation. Instead, I've been looking at mathematical treatments of program/proof syntax, and I've been especially focused on modularity (multi-user eventual interaction, you could say).