I've been trying to resolve homeostasis and plasticity in my head. Those who are interested in homeostasis suggest that neurons have an optimal excitability. Following a perturbation of that parameter, control mechanisms come online and tune the neuron back to a characteristic firing rate. The level of excitability is controlled by ion channels which could be gated by neurotransmitters, voltage, or intracellular signaling. On the other hand, the goal of plasticity is to alter a cell's firing characteristics. For instance, place cells in the hippocampus fire when a rat is in a particular location (the place field for that cell). If place fields remain constant over the course of several recording sessions the place field stability becomes somewhat analogous to a memory.

Some have suggested that homeostasis works in parallel with plasticity to produce a high signal to noise ratio. If excitability is increased at some inputs to a neuron it will be decreased proportionately at others, so the neuron maintains an overall firing rate. This seems at odds with the idea that a memory can be encoded by changes in a cell's firing rate though. To be concrete, say a neuron develops a new stable place field and the rat is confined such that it spends a majority of its time in that location. Will homeostatic mechanisms eventually resolve the cell's firing rate back to the original level? It seems unlikely. It seems more likely to me anyway that when a certain level of plasticity is desired, the signal must be sent to reset the neuronal thermostat to a new desired level. Plasticity without disrupting homeostasis is doomed to sputter out.

An interesting empirical question would be to ask on what timescale the homeostatic detector integrates. How long does it take a neuron in vivo to detect a change in firing rate and return to baseline? It can't be instantaneous or no signals could be encoded via firing rate, but if it takes very long the neuron would become useless for coding any new information as the signal saturates. Some studies have suggested a timescale on the order of 24 to 48 hours; Schuman and co have been able to make one brand of homeostasis visible over the course of about 3 hours. Could cellular homeostasis mechanisms be an explanation for forgetting? Are forgetting curves smooth or marked by peaks at the times that these mechanisms dominate?

Homeostasis reading:
HOMEOSTATIC PLASTICITY IN THE DEVELOPING NERVOUS SYSTEM
Homeostatic control of neural activity: from phenomenology to molecular design.
Homeostatic plasticity and NMDA receptor trafficking.