There’s a “dump heap” hypothesis of agricultural origins which suggests that people first got interested in actively managing and manipulating plants for food or other products when they saw them sprouting out of piles of garbage in and about settlements. There they could observe them daily and experiment with them. A slight variation on this theme — involving corrals in pastoralist campsites rather than garbage dumps — has been proposed for the domestication of quinoa.
One of the things that might have happened in these fertile micro-environments in close proximity to human habitations is that different related species might have been brought accidentally together, leading to hybridization and the development of interesting new — polyploid — types. But there really hasn’t been much empirical evidence for this.
No more. A new paper ((Colin E. Hughes, Rajanikanth Govindarajulu, Ashley Robertson, Denis L. Filer, Stephen A. Harris, and C. Donovan Bailey. Serendipitous backyard hybridization and the origin of crops. PNAS published August 17, 2007, 10.1073/pnas.0702193104.)) looks at the domestication of the legume tree Leucaena in Mexico, where it is grown for food (it is also used as a fodder in some parts of the world). A variety of evidence is discussed which suggests that there has indeed been much hybridization among up to 13 different wild species of Leucaena in Mexican backyards. This has proved “a potent trigger for domestication.” The authors think a similar thing also happened in Mexico with two other perennial crops, Agave and Opuntia.