I consider simulation based science interesting as theory (theoretical science), but it should NEVER be taken as accepted fact.
It depends on the field and the degree of accuracy. For example, hydrodynamic simulations are pretty much accurate for standard(ish) shapes. Same with lighting simulations (ie, ray tracing).
Those are based on experiments. Only after the experiments were the simulations added. And with those it's possible to compare a simulation and a realization and improve the accuracy of the simulation.
i.e. the science is based on experiments, not simulations. After the science was learned, simulations were added, but the science was not learned from the simulation.
It depends on the field and the degree of accuracy. For example, hydrodynamic simulations are pretty much accurate for standard(ish) shapes. Same with lighting simulations (ie, ray tracing).