So many things for conversation here...
On one side, we've got the need to fit into the rules of Formal Logics for our math tools to be useful. Exactly as you pointed out — sometimes information is not sufficient. But a good logician should be able to recognize the holes in the logic behind a problem. Without knowing details. It's either a case where the info covers what it claims it covers; or not. The fallacy is in the claim of sufficient information. And those are things philosophers could be right about thousands of years ago.
On the other hand, there is the usefulness of Statistics — it's a matter of scale. If it's been built on a scale where certain events can cause nothing but ripples on the surface of a larger picture, then it's reliable. If there's anything that can make a ripple as big as a tsunami wave, that will change the landscape, then we have obviously not thought on a big enough scale.
In Isaac Asimov's "Foundation saga" there exists the powerful enough mathematical model Harry Seldon calls Psychohistory (but it's still mostly maths) that predicts events in the galactic civilization for hundreds of years ahead. With all ripples smoothed out in the end.
In a system that is...wait for it...
enough, if there's something to be done, somebody will succeed in doing it, given enough time.
That's how I understand those books.
The question is, what is a truly decentralized system? A galaxy would be relatively decentralized in my opinion. Even a portion of a galaxy could work.
We have the world to work with. And it should be enough for us since we can do nothing about the rest. Yes, we should have a good model, but we must know at all times. It is uncertain, whatever it is.
Anything set in stone lives as long as nothing too big happens to the stone.