The problem of fragmentation prediction is at the origin of various analytical models. Among them, we focus on the ones introducing the idea of obscured zones. They assume that when a crack initiates at a defect, a stress release wave propagates away from the crack and protects the region encompassed by the wave from any further crack initiation. In this paper, we show by the use of numerical simulations that this assumption is only valid at high strain rates. The limit of its accuracy is even pushed to higher strain rates when the fragmentation process becomes more complex, that is to say when crack propagation, bifurcation or coalescence together with wave reflections are implied. In these cases, fragmentation lasts longer than the time needed to completely obscure the whole specimen and the obscured zone theory for fragmentation appears inadequate. We use the cohesive-element method to describe the damage and failure of the material considered.