This particular problem is heavy on game theory, which is a pretty well-developed field at this point. So I think there are grounds to believe we might be able to predict consequences with some accuracy.
And look how it turns out every time that principle is broken, albeit in userland and not the kernel: KDE 4, PulseAudio, Systemd.
I'm fully aware of the differences between a product (airplane) and software, but the 737 Max is born of the same principle being broken. The problem that was being fixed was fuel efficiency. So Boeing replaced the engines and associated structures with more efficient engines, which required different structures to hold them, which required MCAS, and now the whole thing is untested, as compare to the old setup with decades of proven reliability despite 1.6% more fuel consumption.
I'd also point out that arguments like this, while often well-intentioned, can also be abused, by people who don't like a particular new idea but can't articulate a defensible reason. It's easiest then to just fall back on conservatism.
The real answer is to invent new stuff and try not to be idiots about it. Use good design principles, don't cut corners, do the testing, roll out gradually when that's feasible, etc. For the Torvalds principle you mentioned "untested" new code, and Boeing definitely cut corners.
If you are adding complexity into a very complex system, in order to govern it, it is very likely that the change will make the system harder to control. So you must have a pretty good certainty that this will give you governance and not destroy it.
Can you come up with specific ways it might fail? I can definitely think of issues that would have to be addressed, but so far all the critical comments in this subthread could apply, unmodified, to thousands of unrelated ideas, rather than just this one.
> Can you come up with specific ways it might fail?
It's trivially exploitable by the executive team by just behaving in a sustainable way for a while, and then changing it suddenly when every other shareholder is locked-in.
The reason the comments apply to thousands of possible suggestions is that the problems appear on thousands of possible suggestions, and nobody wants to keep writing thousands of comments debunking each bad idea.
Again, it's an extra bonus to the unethical ones, because they are the only ones that can sell, and will take all the profits from the short lived stock pump.
Do you also have a home-made encryption algorithm you'd like to push?
Perhaps you could point me to an equivalent field for making society better, with a well-developed utopia that lots of people live in today.