Reproducibility:

• We identified a problem that the scheduler analysis generated sometimes different numbers of counts in the histograms when we submitted the runs for analysis at different times.

• In addition to examining the .log, .out and .err files, we have also examined the files generated by submitting jobs to the scheduler. Looking particularly in the .condor.log and .list files, we can tell when the analyzer doesn't completely examine a run. We have also examined individual histograms in the ROOT files to confirm that we reproduce the same histograms when the analysis doesn't fail.

• As long as we make the checks and re-submit the runs we have been able to reproduce histogram counts consistently. We are studying what sometimes causes the analysis to abort.

Optimizing fits:

• We have also studied optimal fit window starting points and sizes. (see attachment)
• We have found some characteristic ways in which the code sometimes fits the data poorly. We implement checks and remedies in the code to resolve these characteristic fit problems.

• For non-characteristic fit problems we are looking at manual fits.