My experience is: once you have enough different FSM and reducers that you need to abstract them, the 'sufficiently complicated' criterion is made.