Wire CA Activation Decay

One way to make a CA decay over time is to wire it up so that it decays over time. For example:
If the fatigue constant is bigger than the fatigue recovery constant, say .6 and .1, then one pair will turn off every 2 cycles.
You can set the weights to have them fall out as you like.
Moreover, you can group these with more traditional CAs so that things can be learned and you get all those properties.
However, this just doesn't feel neurobiologically realistic.
I don't really feel that the brain sets aside these little sets of neurons to make CAs decay.
It seems like the behaviour should emerge from the system.