I see small dark and bright regions in my normalized flat field - why?

These features are caused by straylight from 0^th order light in adjacent slitlets. By default the pipeline normalizes the flat fields only along the dispersion axis using a flat field "spectrum" collapsed along the slit. The slit illumination instead is kept (sradius = -1), so that it is present to correct the slit illumination of the science data.

If an adjacent slit has its 0^th order within the valid wavelength range of the slit being normalized, the stray light from this 0^th order peak creates a strong local gradient in the flat field of the slit being normalized. The collapsed flat field "spectrum" contains an averaged value of this gradient, which then creates a hole/peak combination in the normalized flat field. This in turn creates a peak/hole combination in the flat-fielded science data.

The science data will also be contaminated by the 0^th order light, but in a different way as the flat field spectra are all from one source while the science spectra are created by different sources.

This effect can be reduced by setting sradius > 0. In this case the spatial profile is determined by smoothing along the slit, row by row, which corrects most of the effect, but also removes the general slit illumination profile.