A collisional inhomogeneous broadening of the H2 Q(1) line perturbed by Ar was observed for the first time 25 years ago. Several attempts were made to explain the line broadening from ab initio calculations, which however resulted in fundamental discrepancies between the theory and experiment. To resolve this problem we investigate two possible sources of these differences. First, we repeat the ab initio calculations of the broadening and shifting speed dependence, using in the scattering calculations a new, highly-accurate ab initio H2-Ar interaction potential. Second, we replace the previous phenomenological models of velocity-changing collisions with a more physical one based on the interaction potential. This allows us to properly reproduce the experimental broadening over a wide range of temperatures and pressures. We show that this abnormal broadening is caused by strong competition between the velocity-changing collisions and speed-dependent shifting, especially at high pressures.