In order to do that you need some theory (= a collection of mathematical expressions and rules how to use them). Then you can start to test perturbative and non-perturbative approaches. For QCD we have both (applicable in different regimes) and it seems that it works quite well.
As long as we start to write down a Lagrangian which contains gauge bosons we will never find sonmething else and the final theory will always contain gauge bosons; the question whether we do perturbative or non-perturbative calculations is only a technical one.
So the picture with gauge bosons will not break down when we do non-perturbative calculations using gauge bosons; it will break down when we write down something which does not contain gauge bosons ;-)