#1 Think about it. For islands like South Sandwich, solar power isn't an easy deal...... Maybe on Kingman Reef, but not South Sandwich or Kerguelen.
#2 Show me where there is a T1 connection to South Sandwich or Kingman Reef. After all, latency is one of the toughest obstacles to deal with in remote operating.
#3 Good luck with a tower lasting 6 or 12 months there. Hell, stuff at PJ2T and P40 breaks all the time, and that is with not nearly as much wind.
Yes, I think South Sandwich, for the imaginable future, is a bridge too far. Too forbidding. But, as you point out, not every place is that bad.
But, we don't need a T1 connection. Latency is indeed a problem, but it can be managed. A T1 connection isn't required. What's required is that there isn't a lot of competition on the frequency. Maybe somewhere in the 400 or 900 MHz ham bands, or even 2 meters. It depends on where the human sits and what the regulations allow. The added lag for the terrestrial station isn't that big an addition to latency here. Even a thousand miles is nothing; it's making sure there aren't ethernet style collisions.
We only need one or two audio channels to run the whole thing. We don't even need a lot of frequencies here. One could imagine (for argument's sake if nothing else) a remote rig with crystals for frequency control. Point is, the control logic would be low bandwidth. What we need is "enough" audio bandwidth and something like isochronous communications flow. The absolute bit rate is not the challenge.
We don't need a tower, either. For a lot of these places, the Team Vertical approach works very well. A couple of phased vertical dipoles (e.g. one at EU, one at NA and the "other side" hopefully covers the rest) could cover most of the world with adequate gain. The big question would be the power budget. Is 100 watts really enough, even with (say) a 3 or even 5 dB gain from the phased array? (A pair of pairs should work the world in most cases; in extremis
we could go with a single vertical).
In any case, walk before you run. Before we start saddling ourselves with the added complexity of a tower (and its multiple failure modes in the rotor), making a switched vertical array, or even a single vertical, needs to be tried out and tried out in many places.
You might even operate for a week or so on-site before leaving the automated station behind. This would allow it to be debugged and also knock down the hordes to a more manageable size.