yeah, groupcount is sure a part of the "problem".
Also what the groups are doing makes a difference. A T-80 platoon sitting somewhere 'dumb', or them moving across the map, is a different thing.
The more intelligence you add to the map, the faster you are going to run into the problem.
I really would like to know a way of 'calculating' the limits, so I can keep it in account for designing maps that are meant to run for a longer time.
Currently, it's trial and error.