Really interesting project. I see you've dropped the full game theoretical idea? Was this mainly because of the potential blackmail problem derived from the "impossibility" to create equal bargaining power between the two actors which would cause at best a transaction that would would take a rather long time or were there more (technical) problems with your original model? The bargaining problem currently looks like it can only be solved if you could somehow involve characteristics of the bank transaction in the contract.
I'm currently also following the discussion in IRC and as soon as I've got more free time available I'll be able to participate more actively. If developed properly it can definitely have a major positive effect on the bitcoin eco system.
There were some concerns about attack scenarios which cannot be prevented by the original design. It is open if it would work good enough in practice, but I lost confidence and you are confronted with certain risks if you design a project which will not work out well in practice.
My main problems with the game theoretical setup are that the potential loss is asymmetric and that if one party acts unfair the other party lose more. So that will be considered as an unfair setup (game) and lead to "out of the game" solutions (breaking out of an unfair game seems positive for both parties).
I came to that conclusion when listening to a broadcast about a "prisoners dilemma" game setup, where the original game setup is designed against the game players and a breakout is a valid strategy.
Other problems was that its hard to define what is rational, in fact in reality there will be always a mix, specially in blackmail scenarios rational behaviour will be mixed with irrational one.
And system attackers are not covered by the Nash idea. Therefore other solutions are needed (lottery, reputation,...) which introduce more complexity and new problems (Sybil).
At the moment we try to design it with escrows, which will solve a lot of the problems.
Contract, identity and reputation will be added as well, but we need to keep the basic concept as simple as possible, otherwise it will explode to a huge project which will be unrealistic to get finished ever...
Feel free to join, we need for sure more work and brain forces!