Some Quantum Physics research has started yielding results that I think would show up if you were using wholly wrong abstractions/tools/techniques to understand something.
And yet we keep measuring what our models predict to extremely high levels of precision. We know there is something missing (gravity not playing nicely with quantum theory being the most obvious example), but it has worked remarkably well whether you like it or not.
And with the proper insight, we could make our current models look like as crufty as alchemy does today.