Legume agronomic research in Southern Africa has often focused on integrating legumes into smallholder cereal cropping systems, but there is limited information available on the feasibility and soil health implications of continuous cropping legumes in the region. Continuous legumes may be suitable in areas with large livestock populations where a premium is placed on high-quality forage, or where efforts are underway to reclaim degraded cropland. Our objectives in this study were to (i) evaluate the performance of diverse legumes under continuous cropping and conservation tillage management with no fertility inputs and (ii) assess the response of soil health parameters to continuous legumes in a semi-arid environment. A 4-year study was conducted in Limpopo, South Africa beginning in the 2011–2012 growing season in which 10 legume and fallow treatments were imposed in the same plots for 4 growing seasons. All legumes responded negatively in varying degrees to continuous cropping in terms of biomass and nutrient accumulation. Lablab (Lablab purpureus L.) was the top-performing legume in the study and accumulated 4.5–13 Mg ha−1 of biomass and 153–345, 11–34, and 75–286 kg ha−1 of N, P, and K, respectively. Lablab often outperformed natural fallow, while other legumes generally performed as well as or inferior to natural fallow, depending on species and growing season. Cowpea (Vigna unguiculata Walp) was especially incompatible with continuous cropping and averaged less than 252 kg ha−1 and 2.1 Mg ha−1 of grain and biomass, respectively, from 2012–2013 to 2014–2015. Continuous cropping did not lead to sustained improvements in soil health. By 2014–2015, soil organic matter for all treatments had either declined or resembled baseline values. Rates of potentially mineralizable N in cowpea, lablab, vining mucuna (Mucuna pruriens var. Utilis), natural fallow, and bare ground plots fell by 70–96% during the study. There was also evidence for lower recovery of leached K by legumes compared to natural fallow species. In conclusion, legumes, such as lablab, should be considered as continuous forages on marginal land in areas where high-quality forage is in demand, but continuous cropping legumes without fertility inputs are not an effective strategy for improving soil health on degraded cropland in this semi-arid region of Southern Africa. Future research efforts may focus on the grazing strategies and baling frequencies required to optimize annual biomass accumulation of continuous lablab to meet livestock demand and support smallholder livelihoods.