Without getting into too much detail, our process relies on a very delicate instrument to measure a specific product property.  In order to maintain the calibration of the instrument, a known standard is checked weekly to verify calibration and is tracked via a XmR chart.  The known standard degrades over time such that we get only 20 or 30 measurements before it has to be replaced with another known standard.  The result is that each "standard" is slightly different than the previous standard meaning the process is potentially changed with each new standard.

As Dr. Wheeler points out, we can't wait for 30 samples to calculate limits so we calculate them at 3 or 4 samples and then update as we go.  Dr. Wheeler's description of signals and false alarms is right on the money.  The XmR charts behave exactly as described in this article and are very useful.  Our emperical data supports the results of the simulation study very well.  Excellent work!