Scientific theories are theories that predict the behavior of phenomena within the realm in which they have been tested (within some range and some degree of accuracy) and found to agree with the Theory and not falsify it.
Tests that are Outside the realm which has been thoroughly tested (extrapolations, or in some cases interpolations), form falsification tests of the theory.
Future theories would still be constrained to agree with the tested predictions of the current theory (again, within in some margin of error) but could give us a different understanding of the underlying mechanism.
As our knowledge expands, possibilities for Nature that once existed are found to be untenable.
The simplistic, classical world is no longer a viable view of reality - even as it serves as a perfectly serviceable approximation of our common experience.
We already know that Quantum Mechanics and Relativity are not compatible and complete descriptions of reality, but they are extremely accurate MODELS (within nearly every realm which our instruments can currently reach and have held up over many orders of magnitude). We also know a tremendous amount about what any future model CANNOT be (and that guides our explorations forward in terms of M-theory, LQG, holographic models(all but ruled out now), etc.
The scientific methodology is nothing but our very best attempt at removing sources of error, illogic, and biases from our conclusions. Seen in this light, the idea that science must be excluded from some domain of inquiry is laughable because it implies that a methodology that admits error, illogic and bias will give superior answers.
The methodology of faith, meanwhile, has produced a continuous bifurcation of beliefs, resulting in tens of thousands of conflicting and variant belief systems about the very thing it claims it can give us absolute answers about.