A discussion on slack noted that images processed with linearity corrections had amp-to-amp offsets that were not seen in uncorrected images.  The most likely cause was the low-flux end of the correction not being anchored at zero, resulting in the sky levels being set from the extrapolation off the lowest flux value listed in the correction table.  As most of the development work with the spline linearizer was done for PTC improvements at high flux, this hasn't been noticed before now.

The linearity correction calculation should properly fix the lowest-observered flux level to zero, ensure that there's a zero-flux anchor as well, and a similar fix added to the application code to allow existing linearity corrections to be used.