Not 100% sure I understand about the "delta functions". I guess if the edge rolloff is totally monotonic, and quite extreme, then the pixel grid acts to quantise a function that goes from 1 (in the bulk) to (1 minus something significant) in the last row. I guess this is what you mean?

I'm amazed that the field would be so flat as to work like that, but maybe this is the case. If that is the case, the corner amps should look very different with respect to wiggles, as they have a lot more edge (by a factor of 5x). Do we see that to be the case?

I still think that given just how clearly separated these wiggles are (i.e. nice separate peaks on the histograms) it would be really interesting to plot each wiggle, one by one, spatially, to see just how they look. If your theory is correct, it should show up as neat bands, each moving in as you go up in wiggle-index.

I made a simulation of a small flat field with edge distortions: 

With no noise, the histogram shows the structure that looks like a comb function with preferred values: 

Adding Gaussian noise (ignoring Poisson or some other type), the deltas widen, creating wiggles:

Also, coming back to the AuxTel data, this is a zoom in of the wiggles: 

This is one of the peaks (the leftmost), in each amplifier: 

Using amplifier C12 as an example, four peaks of wiggles in the histograms roughly map to the locations near to the lower edge of the detector: 

The four wiggles identified are: 

1): [0, 0.15] ADU/s 

2): [0.15, 3]ADU/s

3): [3, 4]ADU/s

4): [4, 5]ADU/s 

Their location in the pixel array is close to the edge. The intervals separate in bands, providing supporting evidence to the idea that the wiggles in the histogram are preferred values due to the edge distortion effect ("roll-off"): 

1)>4000 ; 2)>8000, 3)> 12000, 4)> 16000 (arbitrary values). 

Very cool! One thing I don't quite get is why the yellow band (wiggle #4) is twice the width of the others, i.e. two rows. At least by eye, the population of that bump in the histogram doesn't looks twice as populous as the others.

A zoom of the edge of that same amplifier (C12) as seen with ds9.