Nick, even were I to agree that you have a counterexample in some particular node of a linked network, it wouldn't matter to the complex system (CX) model. You write:
"My real problem with the general systems and complexity theory approach to just about everything is that it tries to smooth away distinctions in the quest for common patterns. Everything becomes isomorphic."
Network self similarity is not identical to isomorphism. I think you underline here a common misunderstanding of the model -- the power in Bar-Yam's solution to the problem of bounded rationality is the feature I repeatedly quote: "In considering the requirements of multi-scale variety more generally, we can state that for a system to be effective, it must be able to coordinate the right number of components to serve each task, while allowing the independence of other sets of components to perform their respective tasks without binding the actions of one such set to another." That's lateral over hierarchical, cooperative over competitive, and the self similarity of connected maps is continuous at multiple scales of observation.
A self organized system is everywhere both self similar and self limiting. Point is, that network connections over multiple scales are locally time limited and globally continuous. So saying something about time limited local events (your examples) in isolation doesn't say anything about how the events are network-connected. I tried to get this across in my ICCS2007 PowerPoint.