Between plagues, the solitarious desert locust (Schistocerca gregaria) is generally thought to exist as small populations, which are particularly prone to extinction events in arid regions of Africa and Asia. Given the high genetic structuring observed in one geographical area (the Eritrean coast) by former authors, a metapopulation dynamics model involving repeated extinction and colonization events was favoured. In this study, we assessed the validity of a demographic scenario involving temporary populations of the solitarious phase of the desert locust by analysing large-scale population genetic data. We scored 24 microsatellites in 23 solitarious population samples collected over most of the species range during remission. We found very little genetic structuring and little evidence of declining genetic diversity. A Bayesian clustering method distinguished four genetically differentiated units. Three groups were largely consistent with three population samples which had undergone recent bottleneck events. Nevertheless, the last genetically homogeneous unit included all individuals from the remaining 18 population samples and did not show evidence of demographic disequilibrium. An approximate Bayesian computation treatment indicated a large population size for this main genetic group, moderately reduced between plague and remission but still containing tens of thousands of individuals. Our results diverge from the hypothesis of a classical metapopulation dynamics model. They instead support the scenario in which large populations persist in the solitarious phase of the desert locust.