In CPython, @functools.lru_cache would also record the order of the keyword arguments as they are passed to the function to cache the output for. So, a function called with (a=1, b=2) and called with (b=2, a=1) will take up two entries in the LRU cache even though they are virtually the same function invocation.

MicroPython, however, seems to pre-sort keyword arguments. This means the keyword arguments order is lost, and therefore the behaviour of @functools.lru_cache differs slightly between the two implementations.