Meiosis is an ancestral, highly conserved process in eukaryotic life cycles, and for all eukaryotes the shared component of sexual reproduction. The benefits and functions of meiosis, however, are still under discussion, especially considering the costs of meiotic sex. To get a novel view on this old problem, we filter out the most conserved elements of meiosis itself by reviewing the various modifications and alterations of modes of reproduction. Our rationale is that the indispensable steps of meiosis for viability of offspring would be maintained by strong selection, while dispensable steps would be variable. We review evolutionary origin and processes in normal meiosis, restitutional meiosis, polyploidization and the alterations of meiosis in forms of uniparental reproduction (apomixis, apomictic parthenogenesis, automixis, selfing) with a focus on plants and animals. This overview suggests that homologue pairing, double-strand break formation and homologous recombinational repair at prophase I are the least dispensable elements, and they are more likely optimized for repair of oxidative DNA damage rather than for recombination. Segregation, ploidy reduction and also a biparental genome contribution can be skipped for many generations. The evidence supports the theory that the primary function of meiosis is DNA restoration rather than recombination.