Think of it like Geometric level of detail in video games: when the observer is viewing a character from very far away a low-polygon mesh is displayed -- this is because the mesh only takes up a few pixels of screen space for the observer so rendering a high polygon mesh or a low polygon mesh is indistinguishable.
Now when the character is closer to the observer a high polygon mesh with more detail is substituted.
The observer can never tell that this trick is happening. If they designed an experiment to test for this the universe could simply spend more time fooling the observer.
I am quite sure if we are living in a simulation it makes a lot of approximations. When we conduct complex experiments it simply computes better approximations so that our observed results fit the "axioms of physics".