It stagnated when they started doing NHST, ie checking for a difference from "background" vs collecting and comparing data to the predictions of various theories to distinguish between them. Same thing that has destroyed every field of research that adopted this approach.
Imagine if Einstein just predicted that the position of the stars would appear to be different during the eclipse, rather than displaced by an exact amount. The last 40 years has seen physics become more like the former (bad) than the latter (good).
