Experimental evolution is a powerful approach to study the process of adaptation to new environments, including the colonization of eukaryotic hosts. Facultative endosymbionts, including pathogens and mutualists, face changing and spatially structured environments during the symbiotic process, which impose diverse selection pressures. Here, we provide evidence that different selection regimes, involving different times spent in the plant environment, can result in either intra- or extracellular symbiotic adaptations. In previous work, we introduced the symbiotic plasmid of Cupriavidus taiwanensis, the rhizobial symbiont of Mimosa pudica, into the phytopathogen Ralstonia solanacearum and selected three variants able to form root nodules on M. pudica, two (CBM212 and CBM349) being able to rudimentarily infect nodule cells and the third one (CBM356) only capable of extracellular infection of nodules. Each nodulating ancestor was further challenged to evolve using serial ex planta-in planta cycles of either 21 (three short-cycle lineages) or 42 days (three long-cycle lineages). In this study, we compared the phenotype of the 18 final evolved clones. Evolution through short and long cycles resulted in similar adaptive paths on lineages deriving from the two intracellularly infectious ancestors, CBM212 and CBM349. In contrast, only short cycles allowed a stable acquisition of intracellular infection in lineages deriving from the extracellularly infecting ancestor, CBM356. Long cycles, instead, favoured improvement of extracellular infection. Our work highlights the importance of the selection regime in shaping desired traits during host-mediated selection experiments.