The internal feedback hydrostatic rotary table is a precision supporting device, the performance of which largely depends on the oil pad, a core component. However, uncertainties in the manufacturing process are often overlooked during the stiffness optimization, affecting the reliability of the optimized results. Accordingly, this paper aims to analyze the influence of structural parameters on the stiffness performance of the internal feedback hydrostatic rotary table and to perform reliability optimization considering the uncertainties. Initially, a theoretical computational model of internal feedback hydrostatic rotary table, accounting for the oil leakage effect, is proposed. The model's accuracy is validated through comparative simulation calculations, and based on this model, the load-bearing performance of the table is further analyzed. Subsequently, focusing on the structural characteristics of the oil pad, a reliability optimization model that considers manufacturing uncertainties is proposed. To improve the optimization efficiency, a Levenberg-Marquardt Backpropagation (LM-BP) neural network is introduced as a surrogate model for theoretical calculations. The oil pad is optimized through a particle swarm optimization algorithm. Ultimately, the optimal structural size parameters of the oil pad are obtained, achieving maximal stiffness under a high level of reliability. Both the stiffness performance and the reliability level of the rotary table are substantially enhanced. The results indicate that the proposed method significantly improves performance and reliability in practical applications.