You are also postulating a discrete model. There is a computational parallel in the neural network area, it's called "liquid state machine", it's worthy of study. There are "phases of matter" in a neural network too, they're defined computationally. Simple networks operate like ferromagnets that lose their alignment once they reach the Curie temperature. Then they become "liquid", all the little dipoles point in different directions.
You can decide if that's helpful information. It seems to me, that maintaining a "structure" (solid phase) would be the normal condition, a liquid phase would change the rules for transmission from one cell to the next.