I think you're underestimating the signal strength side of the problem.  Simply *detecting the presence* of an EME signal is a difficult step.

Max Power wrote:

================
In Europe there is a PAL system that uses only 5.0 MHz channel capacity.
NTSC-BSTC uses (technically) only 4.58 for vision, and comes in under 5.85 MHz ignoring 'guard bands'.

You can receive a stable and identifiable (analog) picture with considerably less than 4MHz - I'd bet 1MHz would be enough bandwidth if the identification slide was large enough.  Still, that's about four orders of magnitude wider than what the amateur experiments seem to be succeeding with.

But when you go to ATSC digital you need the whole channel.  (there's a lot I don't know abut DVB-T but I strongly suspect it requires the whole channel as well)

I would not expect the PAL signal to survive the trip intact, however.
Perhaps using NICAM to recover probabalistic sync intervals might work, but recovery of the coluour signal (due to doppler and ionsphereic distortions) I would classify as theoretically impossible -- unless you are using AIRICEBO / CANBERRA / GLADSTONE.
A detailed systems analysis is needed to verify most of my assumptions, as probably a few are wrong.

Given the attenuation issues I don't think many people have considered what doppler, ionospheric distortions, and reflection irregularities would do.  I would imagine they would make things pretty difficult for ATSC.  Reception of analog signals via F-layer propagation usually exhibits severe multipath, more than enough to render F-layer ATSC reception almost certainly impossible.  (E-layer, on the other hand, has already happened)

I suppose the Arecibo dish may in fact be large enough to deliver enough signal for NTSC analog reception.

Then, there's the issue Paul K. brings up: interference from multiple stations on the same channel.  The amateur experimenters are avoiding this by using extremely narrow receiver bandwidths and predicting the Doppler shift characteristics of specific target stations.  This works fine for resolving the presence or absence of a carrier.  Much less so once you open the bandwidth up to 1MHz or more and start having multiple signals interleaving in the same spectrum!