interesting test.
I found about 15 bright lines of 768 lines for 1/4000 s:

http://www.fho-emden.de/~hoffmann/flicker20122006.pdf

The doc shows as well the frequency response for the
pupil control system. No interpretation in the moment.

I'm expecting damage mainly because of aliasing effects.
A confusion between flicker frequency and sampling /
transmission speed by eyes + nerves.

When looking at a CRT the eye is subjected to something like 5000cd/m2
flicker that will average to something like 100cd/m2. This CRT flicker
is so very rapid that the pupil reflex (in the usual/common sense) can
not be activated due to it.

So, when we look at a paper that has 100cd/m2 (illuminated by direct
current lamp)  then the pupil is certainly adjusted naturally, it will
be wide open or close to that.

But when we look at a CRT that has 5000cd/m2 flicker that averages to
100cd/m2 then how is the pupil adjusted in this case? In case it is
wide open similarly as above then there will be somewhat higher load to
the optic nerves. It would also be an un-natural case that possibly
could affect to the way we preceive the tonal range (especially in the
dark end) on the CRT compared to the same situation on the paper.

the Web offers little about this issue. 5000cd/m2 is rather bright.
Can we interprete this as the luminance of a certain bulb or tube ?
The linear averaging (here to 100cd/m2)  seems to be correct,
see the chapter about flicker in W+S.

This is a Swedish doc:
http://epubl.ltu.se/1404-5494/2006/015/LTU-HIP-EX-06015-SE.pdf

Ref 19 concerns a publication by Berman, Greenhouse and Bailey.
This publication is often quoted but not available on-line.
The authors had proved, that flicker is transmitted by eye+brain
for frequencies above the Critical Flicker Frequency or Fusion
Frequency (max. 45Hz), up to more than 142 Hz.

White paper under sunlight can have far higher luminance.

 > The linear averaging (here to 100cd/m2)  seems to be correct,
 > see the chapter about flicker in W+S.

Indeed, so why is it that the alleged non-linearity of the tonal range
of the vision does not come into play at all in this situation?

I'm understanding your point: luminance vision is linear if the
adaptation state is fixed.
But the flicker test is perhaps useless as a proof - if the
averaging happens by optical-chemical processes instead
of neuronal 'calculations'.

About the lightness of the horizontal bar on a CRT monitor:
maybe 5000 cd/m2 for an averaged luminance of 100 cd/m2.
White clouds on a sunny sky: 6700 cd/m2, blue sky 4000 cd/m2
according to René Bouillot, La pratique du reflex argentique
& numerique.
But this is misleading, because eye+brain are adapted to the
average monitor luminance.

The next part is a simplified description for a test, using a
SoLux bulb 4700K and the matt side of Chromolux carton,
probably a Lambertian reflector.

L = (1/pi) R E cos(Phi)

L: Luminance in cd/m2
E: Illuminance in lx
R: Reflectance factor, perhaps 0.8
Phi: Angle between light direction and reflector normal, zero

Monitor averaged:
E = 400 lx
L = 100 cd/m2

Monitor peak:
E = 20000 lx
L = 5000 cd/m2

The SoLux bulb delivers approximately E1 = 1000 lx in distance
d1 = 1m, as measured by Spectrocam for the reflected light.

For E2 = 20000 lx the distance d2 is calculated by
d2/d1 =Sqrt(L1/L2), which delivers d2 = 0.22m .
This spot is perceived as 'very bright' for monitor adapted eyes.