Depends how its set up. So long as it's fully independent and disconnected from existing digital infrastructure it should be safer. It could be as simple as explosives hard-wired with a buried line running up into some bunker up in the mountains.

Geopolitics aside, the technical architecture implementation of this mechanism is really interesting for me. I think over all, having extra ability to disable these systems would prevent US launching attacks against the plants — which could cause spill over local civilian injuries — but there’s just so many more things to consider.

Is it a dead-man switch style of setup, where if it doesn’t get authorization from HQ after some time, it will stop working? Or is it a kill switch style of setup, where they can remotely issue a command to stop operation? Because different vectors then come up depending on the securing method. For example: Dead-man switch might be tricked/overcame by turning back the clock, whereas kill switch might be circumvented by severing the network connection before the command could be issued (literally cut the underwater cables before they start the invasion).

How is the mechanism itself secured? If it is certificate based like everything else, then we’d have to worry about the certificate signing authority getting pressured into signing certificates by state backed actors.

Would really love to learn about the setup one day after all these is over, to learn about the thinkings that’s been done on such an important piece of … “infrastructure”?

They'd have everything to lose. Everyone wants those machines. Disabling or destroying those machines is like slashing the only nice life raft on the open ocean. Sure, there are others, but they have cracked rubber and don't seem as firm. Bleeding edge fabs are the oil of the 21st century.