When there were small ranges of values, e.g. in bias images, a lack of consistency in binning logic caused the median algorithm to fail.

Median values will be the same for most images.
Some images had median calculation failures because of an off-by-one error caused by an inconsistency on bin assignment in histogram computations
For instance, in https://indilib.org/forum/ekos/6520-median-computation-in-fitsviewer-wrong.html#50100 Herrhausen provided an image which previously
got median values of 0. With this fix, the correct median values (768 for each channel) should be calculated.