If there is a difference, it is small enough, trivial in comparison to other performance hogs. Unless you are trying to simulate a neural web or something.
Just one thing to consider - the conditions loop through available instances every time they're run. If there are a LOT of instances, it may be better to filter them into categories, so only relevant sets are available to conditions. I sometimes do that when dealing with a huge amount of objects. Set conditions into sub-events where objects are pre-filtered already, so not to re-pick them again. No clue if that helps at all, it sure helps me organize them in my mind, though!
Come to think of it, if this concerns you so much, why don't you test it? Two test capx, different methods in each, see performance impact when dealing with 1000, 10000 and even 100000 instances.