Great suggestions, thanks! Indeed, that's a beautiful connection and many proofs of reinforcement learning algorithms with function approximators actually use contraction mappings.

Say hello to Kwabena for me if you get the chance ;)