> Can you avoid running into power limits that would greatly slow the initial exponential progress?

I doubt it, for all the reasons you give, though I agree the first choice is to do this as part of building a Dyson swarm.

I don’t even know if the interior of the moon is cool enough right now even without the extra heat from manufacturing; but then, I wouldn’t expect to know — I only have a GCSE grade B in chemistry, so I can be sure whoever does invent a von Neumann mechanism, it won’t be me.