If the model isn't linear then presumably it's pretty much unstable, almost by definition. I suppose a subset of small problems might convege towards something resembling equilibrium. But won't a reality-sized model that assumes time-based relationships between inputs and outputs wobble chaotically all over the place?
I agree, though, that as soon as you introduce time and, more importantly, delays, you get an entirely different order of complexity. Delayed feedback is really, really nasty in that respect. Those whom the Gods wish to destroy They first make mad. -- Euripides
