I found when trying to run the parallel overscan in overscan.py, that most parallel overscan values were being masked out. I plotted a histogram of the values, and they look reasonable, and only a few should be masked out. See : Parallel_Overscan_Values_Amp_C10_16Nov22.pdf
The problem is that at this line:
https://github.com/lsst/ip_isr/blob/e3fba828a5a18f981ebfd0a6426b864e7888fb3f/python/lsst/ip/isr/overscan.py#L221
The threshold is treated as a number of sigma, but at this line:
https://github.com/lsst/ip_isr/blob/e3fba828a5a18f981ebfd0a6426b864e7888fb3f/python/lsst/ip/isr/isrFunctions.py#L126
the threshold is a scalar value. When I set isrConfig.overscan.numSigmaClip=30.0, then it behaved as it should. I don't know whether the right fix is to change the name of numSigmaClip or to re-write makeThresholdMask.