My personal feeling is that the "wrong turn" was made when theorists stopped looking at experimental data, and tried to go far beyond what was experimentally hinted at.
It never happened before that theories "came out of the blue" with no experimental hints.
Newton was based upon Kepler's observations. Maxwell was based upon countless empirical laws. Special relativity was based upon a reinterpretation of the symmetry group of the Maxwell equations (the Lorentz transformations) which had already some empirical success.
The closest hit to "a theory out of the blue" was general relativity, but apart from the mathematical difficulty, there was a basic guiding principle and it was "sufficient" to work this out. And even there, there WAS some empirical suggestion (the perihelion shift of Mercury).
Quantum theory was "shoved down the throat" of theorists from spectroscopy (Balmer series and all that). QFT had a difficult emerging, and was essentially guided by some empirical data like the Lamb shift. The entire standard model was a theoretical fit on a huge amount of experimental data, where concepts were introduced in order to explain phenomenologically observed regularities.
And then, theorists left off, and went for a "beauty contest". They invented unified groups (because U(1) x SU(2) x SU(3) was "ugly"), they invented supersymmetry, they invented strings...
It is exactly at the point where theorists didn't try anymore to model experimental data, but where they entered in beauty contests independent of experiment, that wrong predictions came out as indicated here before.
