The molecular clock has proved to be extremely valuable in placing timescales on evolutionary events that would otherwise be difficult to date. However, debate has arisen about the considerable disparities between molecular and palaeontological or archaeological dates, and about the remarkably high mutation rates inferred in pedigree studies. We argue that these debates can be largely resolved by reference to the "time dependency of molecular rates", a recent hypothesis positing that short-term mutation rates and long-term substitution rates are related by a monotonic decline from the former to the latter. Accordingly, the extrapolation of rates across different timescales will result in invalid date estimates. We examine the impact of this hypothesis with respect to various fields, including human evolution, animal domestication and conservation genetics. We conclude that many studies involving recent divergence events will need to be reconsidered.