I wonder whether they've tried any of the optimising emitters in MicroPython:

https://docs.micropython.org/en/latest/reference/speed_pytho...