In this work we address the dynamics of Markovian systems by tracking the evolution of the probability distribution, utilizing mean first passage time theory to augment the set of states considered. The method is validated on a lattice system and is applied, in conjunction with landscape analysis (saddle point searches) and multidimensional transition-state theory, to an atomistic model of glassy atactic polystyrene, in order to follow its time evolution over more than ten orders of magnitude on the time scale, from less than 10(-15) up to 10(-5) s. Frequencies extracted from the eigenvalues of the rate constant matrix are in favorable agreement with experimental measurements of subglass relaxation transitions at 250 K.