From reading previous postings, Mark had excellent reproducibility even when he redid tests on different days, so most likely the error bars will be much smaller than you think. It is cheap and easy to ask for multiple test runs, but expensive and tedious to actually do them.

With that in mind, I would recommend he repeats tests where he measures anomalies in order to make sure that we don't try to solve a mystery which turns out a measurement artifact. Such a repeat test could also tell us whether the anomaly is consistent or random.