Yes I agree with Jousef. Checking your numerics, you have adjusted it to give the simulation the best chance of converging so the error isn’t there nor in the mesh. What Jousef mentioned is quite spot on where your boundary conditions may be incorrect. What you’ve assigned now makes sense in theory because we are trying to represent the environment as “free air”. However, from my guesses, the way you’ve assigned it will cause the simulation to “force” a solution since your outlet is defined as a velocity outlet. On top of this, having the sides as an inlet-outlet as well will cause some of the numerical solution to exit the domain by the sides due to the turbulence effects present in the simulation and coupled with this “forcing” of the solution, your simulation is bound to hit some sort of continuity error at worse and at best, divergence.
It is clear from your first simulation that that the solution is already unstable and diverging as you can see from not only the convergence plots but also the force plots. In the second simulation with the rotating wheel, the two issues I’ve mentioned earlier are even more prevalent, exacerbated by increased turbulence due to the rotating wheels, hence the very absurd (and diverging) values you have witnessed.
Adjusting the side walls to slip, having a velocity inlet defined while the outlet is a gauge pressure zero outlet will likely resolve your issue. If you have concerns or do see the side walls interacting with the results, then you can increase the size of the bounding box both laterally and vertically to negate these effects. I also recommend increasing the simulation end time to 2000s or more and to check the force plots to see if steady-state convergence is obtained where the final force values do not fluctuate anymore.
Hope this helps.