It seems Phoenix will continue to process stale work if new work is not available and old work still is.  My stale rate went up massively overnight compared to when I was mining on deepbit.  One miner was nearly 9%.  One of my machines runs poclbm and the many of the idles were too short to cause the CPU usage and temperature to drop, but not all ... in fact, there were enough that there was always one showing on my graph in MSI Afterburner.  I had to ditch the pool until this is correct as my hardware is too important to fatigue it unnecessarily.