Nobody's competing with me for top scores, which is a little unfortunate - that would be more fun, but it's still interesting to compete with myself and see how high it's possible to go. Apparently Tomf's code uses 1000 as the visited value for walls (just picking an arbitrary big number he thought nobody would ever reach), so if I can get the mouse to visit a tile more than 1000 times, it will power up and start walking through walls. This is my goal: break his assumptions and break the game. I'm only in the 600s at the moment, but there's some vaguely exponential behaviour which causes it to sometimes increase in fairly large steps, so I may not be too far off:
In this configuration, if the mouse is at C and has visited A N times, then it must visit each of B and D N times before it will return to A, and in the process it will visit C 2N times.
Edit: I ran a maze that should achieve 642 visits to the same tile, but at 201 the mouse levelled up and started crossing walls. So I'd gotten the value wrong; I've broken the game already. Awesome. I feel a sense of achievement - I now have the actual maximum possible score given the current implementation. Am now hoping Tomf will fix it so I can have a higher score.