Chemero's reply to the C-C Fallacy

Chemero, 2009, pp. 31-32, claims that the C-C fallacy is only a fallacy when the coupling between agent and environment is linear.  "When the agent and environment are nonlinearly coupling, they, together, constitute a non-decomposable system, and when this is the case, the coupling-constitution fallacy is not a fallacy."

By "non-decomposable" Chemero means
"Nondecomposable, nonlinear systems can only be characterized using global collective variables and/or order parameters, variables or parameters of the system that summarize the behavior of the system's components".

Now, as mentioned in an earlier post, we need to distinguish an extended cognitive systems hypothesis from an extended cognitive processes hypothesis.

I know that it is common to say that "non-linear coupled dynamical systems are non-decomposable", but I'm not sure what that really means or why that matters for the c-c fallacy.  For example, take a double pendulum (somewhat different than was discussed in Bounds, p. 108f.)   You get elaborate equations of motion for this system.  So why is this non-decomposable and why does this mean that you cannot distinguish the two coupled things?  Why can we not distinguish the swinging process in the one pendulum from the swinging process in the other?

Chemero promises a more detailed discussion in other works, so maybe there is an answer there.