Crop yield stability and soil mineral nitrogen (Nmin) have rarely been evaluated from a long-term perspective in the extremely arid cropland regions of China. Therefore, a nationwide experiment aimed to optimize fertilizer application and increase productivity and nitrogen use efficiency in gray desert soils was initiated in 1990. Eight combinations of chemical fertilizers (CK, N, NK, NP, and NPK), straw return (NPKS), and manure amendments (NPKM and NPKM+) were tested for 24 years on spring wheat, winter wheat, and maize. The results displayed that the yield of three crops from balanced fertilizer treatments (NPK, NPKS, NPKM, and NPKM+) did not differ significantly after 24 years; however, reliable yield stability due to lower coefficient of variation (CV) and higher nitrogen harvest index (NHI) were recorded for manure amendment treatments. Compared to NPKM, NHI was lower for the NPKM+ treatment, but crop yield and stability did not improve, suggesting that the appropriate choice for manure amendment is important for guaranteeing food security in extremely arid regions. Balanced fertilizer treatments resulted in lower Nmin residual in the 300 cm soil profile, compared to unbalanced fertilizer treatments. The NPKS treatment gave the lowest value. In the 0–100 cm soil profile, Nmin was higher in NPKM than in the NPK treatment, suggesting that straw or manure amendment can effectively maintain Nmin in the topsoil undercurrent cropland management in arid areas. The NPKM treatment had the highest crop nitrogen recovery rate and the lowest nitrogen losses, further illustrating that manure amendment has higher N retention potential. Overall, although Nmin residues are relatively high in these regions, balanced fertilizer treatments, especially NPKM and NPKS, are the optimum strategies in extremely arid regions.