One or more registers might be getting initialized differently in the emulation than in the actual games, or the game might read an address to seed the random # generator and that address returns a different value in the actual games. In the same vein, the hack to make arcade Ms Pac ROMs run on a Pac Man PCB makes the normal patterns not work because the random # generator uses the ROM contents.
If it was random, I could probably chalk it up to any number of things... But these patterns actually seem pre-programmed, both Pac-Man and the ghosts move _exactly_ the same way _every_ time... Unless the 'random' pattern is seeded from a CPU clock/cycle or something. Since it starts up the same way every time (from off with no memory, and you can't delay the start of the demo), that might explain it. And MESS just getting a different start for some reason. At least they both duplicate their patterns consistently. Just thought it was weird.
For actual games, you leave the system on for a random amount of time before starting... Maybe that does explain it.
