Just a very quick status update. I realized with this chip that it doesn't actually contain an internal quartz crystal, only the capability to use an external crystal without additional circuitry. So you either need a crystal, or drive it from the Gameboy's clock. This should work fine and this is just an aside.
I finally sat down with this the other night, and it failed completely. I still have some ideas for what the problem is, from scope output just before I left the lab. (In particular the behavior of the Gameboy's /CS line as well as a peculiar behavior where the FM chip outputs data to the data bus, while other wise being read only.) I still believe I should be able to get this working with a few changes to my ugly air wire circuit.
Then there's the issue of whether to route the audio through the input, or mix it directly with the "prosund" after the CPU. The FM chip has two outputs, one for tone audio and one for drum audio. One possible configuration is drums as mono, mixed after the CPU, and the tone output connected to the audio input of the CPU, which would allow for panning of it. And I will have to investigate which voltage levels the input can accept.
So, nothing too exciting yet, just a quick status update.