Here are the amplitude errors for before the fix (left) and after (right). These amplitude errors are not calculated in the same way as the chi square value (which is infered from a least squares sum method) but actually from the database recorded error of each strip. This, in term, is determined from the standard deviation of the noise peak in each strip that was inserted in the database. As you can see, the left (unfixed) is rahter constant while the right (fixed) drifts. The rather constant nature of the online baselines is likely due to them being very rarely updated compared to the scale needed to actually make the chi square probability correct.