It is not exactly a new revelation, if you include a model in the model, the problem potentially arise. What you typically do is to keep running the model with the previous answer, and hope the answer converge. It is a common technique for finding numeric solution to the more complex equation systems. I have done it many times. Of course, there is in general no guarantee that there is a solution, that the solution is unique, or that you will find it using this technique.
This all depends on the fact that the simulation and the real world are deterministic which, it would seem, is highly improbable.
This doesn't change much, the problem irons its self out. I see no paradox. And I don't know what you see that makes determinism "seem improbable."
In this argument, every simulation would look into the future to see the lottery numbers, and would act out on this knowledge. There would be no "changing the future." Every simulation would see the lottery numbers, then see themselves go to the store and get a ticket with the numbers, and every simulation will win with those numbers. There was no other future. The future you see in the simulation is the future that will occur in your own simulation.
Calling this a paradox is like calling recursive mirrors a paradox.
It has nothing to do with time travel, if you didn't care to read the article, why care to comment.
Also you can probably only know the future of different, possible similar timelines. Whatever changes you make in those other timelines will not affect your own. You can alter their future but not your own, otherwise you can alter causality. I don't know if altering causality makes any sense.