Unless a hobbiest has very deep pockets, I wouldn't really see this being
done.  Not at the suggested resolution anyway.

The "Jumbo-Tron" used at Toronto Skydome (I think or is it the one in times
square), is actually done with 5" crt as elements/pixels.  One can just
imagine the power consumption.  So power is not necessarily the limiting
factor.

I tend to agree with the "sub-picture" idea.  Feed your video signal
commonly to all subs and let each sub capture the portion of the signal that
they need.  A little manipulation at each element will produce the desired
results.  There would be no need for a common processor with outrageous
computational and memory requirements.

I realize the goal is to make a fairly high resolution image, but unless you
plan to replace the times square display, you would probably not get the
effect of high resolution at such close range.  If you can afford to put
this thing in such a spot, then the cost of building it won't much matter
either.

Just some of my random thoughts :)

John