wow!  good analysis!

looking at your plots, "airglow" is clearly a small component in the
residual noise plots.  i found earlier, using a detailed analysis of scans
definitely classified as airglow from the quality analysts, that:

----------
Large H residual noise is always accompanied by large J and K residual
noise. Plots of J and K residual noise vs. H residual noise for all coadds
with H residual coadd noise above 2 DN and for all values for the scans
identified as having significant airglow show that: 

There is a tight correlation between the residual coadd noise in all
bands when the H residual coadd noise is large. There are occasional
discrepant points where the J or K residual coadd noise is even larger than
expected from the tight relationship. An additional noise source, such as a
bright star, must be present in those coadds. 

The relationship between the J : H : K residual coadd noises is 0.05 :
1.00 : 0.18, significantly different from the 0.13-0.20 : 1.00 : 0.35
relationship in the absolute background level. 
-----------
http://spider.ipac.caltech.edu/staff/tchester/2mass/analysis/galaxies/backgrounds/analysis.html

see the plots referenced there for the background for those numbers.  in
particular:

http://spider.ipac.caltech.edu/staff/tchester/2mass/analysis/galaxies/backgrounds/low_density/colors_excess_noise.html

shows plots where there is clear correlation between the residual
background sigma between the bands.

converting the relations that i previously found for airglow scans:

j residual noise = 0.05 h residual noise
k = 0.18 h

to your score, i get:

j score = 0.09 h score
k score = 0.33 h score

equivalent to:

j = 0.27 k score
h = 3 k score

so only the points along the bottom envelope for your score j vs score k
plot, and along the top envelope for your score h vs score k plot, are
probably due to airglow.  the vast number of other points away from the
envelope must be due to bright stars, galaxies, etc., that you have
consistently claimed cause a lot of the deviations.  (recall that in my
plots referenced above, i just used scans identified as airglow scans, so
the airglow relation stands out in my analysis, whereas your plots are
representative of the entire data set.)

as you say:

We have not seen the last of this issue.  Noise is everywhere and
relentless.

i'll add:

and tremendously complicated....