Dioecy, the coexistence of separate male and female individuals in a population, is a rare but phylogenetically widespread sexual system in flowering plants. While research has concentrated on why and how dioecy evolves from hermaphroditism, the question of why dioecy is rare, despite repeated transitions to it, has received much less attention. Previous phylogenetic and theoretical studies have suggested that dioecy might be an evolutionary dead end. However, recent research indicates that the phylogenetic support for this hypothesis is attributable to a methodological bias and that there is no evidence for reduced diversification in dioecious angiosperms. The relative rarity of dioecy thus remains a puzzle. Here, we review evidence for the hypothesis that dioecy might be rare not because it is an evolutionary dead end, but rather because it easily reverts to hermaphroditism. We review what is known about transitions between hermaphroditism and dioecy, and conclude that there is an important need to consider more widely the possibility of transitions away from dioecy, both from an empirical and a theoretical point of view, and by combining tools from molecular evolution and insights from ecology.