We continue to ratchet up the silly. This time, I loaded five NO-OPs into the SRAM by hand, and then an RST 0, which causes the Z80 to jump back to PC=0000, and we start over!
One thing I didn't realize until right after I finished recording is that the RST 0 command actually is a function call, so the strange address bus activity and writing happening during that opcode is the Z80 pushing the current PC onto the stack, expecting us to then eventually run a RET return to pop it off again...