Self-incompatibility in plants of the Brassicaceae family is controlled by a highly diversified molecular lock-and-key system consisting of a large set of specific haplotypic combinations of only two genes. This system has been a textbook example of natural (balancing) selection, in the form of a strong reproductive advantage for individuals expressing rare alleles. These haplotypes also form a striking linear dominance/recessivity hierarchy, whereby most heterozygote combinations express only one self-incompatibility specificity at the phenotypic level. In this seminar, I will detail how we recently identified the molecular determinants of this dominance hierarchy and showed that it is based on a complex regulatory network based on the interactions between a dedicted set of small non-coding RNAs produced by dominant alleles and their target sites in recessive alleles. I will review several key features of the topology of these interactions and combine theoretical modelling and functional approaches to discuss our current understanding of the functional and selective constraints on the evolution of the network.