The first implementation of collisions which I'm actually satisfied with, or at least I don't see any problems with it yet. Even if this turns out to be the best I can do, I feel it's probably enough to make a game with.

Movement is essentially processed pixel by pixel. It's an expensive approach which needs a serious optimisation pass but I think it might be worth using for the player character. How I convinced myself to use this approach is a long story but basically I think it leads to more consistency which is what I'm after. With any luck some optimisation will make me more comfortable about the performance cost.