You don't get it: it's NOT a matter od "degree" and "resolution".

It's that the model will DIVERGE COMPLETELY in a way that you cannot predict because you can never fully take in account everything.

So a resolution increase of eg 100x will NOT help as the model will STILL DIVERGE.

Even if you take in account 100x more variables in a double pendulum or decrease the timestep by 1000x it will STILL DIVERGE sooner or later.