I think, too, that ARM would be an interesting platform for TCL and would be a nice revival for my Beagleboard. But the problem compared to the other platforms mentioned is, that they have a defined build structure, so everyone can rebuild the binary packages on their own.
In TCL this is not the case. I'm not sure about the base, but for the extensions only binary versions exist. It's up to every extension-maintainer on how to compile and build the extension. I've generated my own script collection for these tasks, as mentioned
here.
So I'm drawing the conclusion, that currently an ARM port could be done as a proof-of-concept-hack. But to get an ARM version, that's as usable as the x86 one, you'll need to redefine the build of the distribution and it's extensions in a more automated way.
If you want to dive into ARM development right now in a very modular and small-footprint way, you might want to take a look at
Ångström and
OpenEmbedded.