Executive Summary
A bug in the calculation of nmip causing it to be an integer occurred starting as of run 20348050.  Fortunately the ADC is fine, so this simply needs to be fixed for the production (and the rest of the run).  This effect is seen in the MuDst.

Details
Takafumi has noticed a difference in the nMIP total from the EPD, which is detailed in his presentation at:  drupal.star.bnl.gov/STAR/system/files/run20detQA_1219.pdf

The key plots are shown below.

Figure 1: One the left is the averagenumber of hits in the EPD versus run number, where a hit is defined as nHit = nMip for 0.3 < nMip < 3 and nHit = 3 for nMip > 3 and nHit = 0 for nMip < 0.3.  The diagonal line by run index 50 is the result of the bias scan.  At run index a little greater than 100, we see a sudden jump in the average number of hits, and this corresponds to a much less stable epd flow vectors on the right.

To be more specific, the Bias voltage scan was in day 345.  The sudden jump is in run 20348050.  This also happens to be around the time that the cool tiles were noticed, this may be followed up at: (drupal.star.bnl.gov/STAR/blog/rjreed/epd-cold-tiles-again  Run 20348027, for example, shows holes in a few tiles).

I have analyzed a subset of the PicoDsts that are available, using all triggers and making a cut of |Vz| < 70 cm.  It turns out that at least the apparent cause was very easy to see.

Figure 2: On the left is the sum total of the EPD hits vs event from before (black) and after (after) the change.  I see the same shift that Takafumi did, and in fact one can see right away that we have float that is somehow an integer.  The middle are the N_{mip} distributions from EW 0 PP 1 TT 1.  It's pretty clear hear that the distribution went strange.  Fortunately, what is shown on the right are the ADC spectra, which look identical and indicate that the data itself is probably fine.

Next is to figure out what happened.  This effect is also seen at the MuDst level.

Figure 3: The same two run numbers as in Figure 2, but this analysis came from the MuDsts, which shows that the problem didn't occur in the production of the picoDsts from the MuDsts.