There could be an inlet for threshold time, and if the trigger inlet received two 0-nonzero transitions within the threshold time the output could jump to a random step........orrr if the trigger inlet DIDN'T receive two 0-nonzero transitions within the threshold time it would jump to a random step. So any slight deviation from the set "rate" would cause randomness.....
Or it could accept a list (eg. the output of a multislider - range of 0. to 1. with the same number of sliders as steps). Each slider could set the probability (for each step) of triggering a jump to a random step.
I might have some more ideas as well, but not sure if they make any sense yet :P
Optimistic plan for the summer = start learning how to make externals