With the introduction of smart manufacturing and green manufacturing strategies, the demand for small batch customized production is increasing in actual production. Moreover, rework production lines are being used more frequently in actual production in order to further improve the processing efficiency and product quality. Based on the context of small batch production, the transient performance of a multi-rework system with unreliable machines and finite buffers is analysed. Meanwhile, the production scheduling of the multi-rework system is carried out. To quantitatively analyse the transient performance of this production system, a mathematical model of the system is established using the Markov analysis method. Besides, a dynamic decomposition and aggregation algorithm is proposed to predict the transient performance indicators such as productivity, consumption rate and work-in-process inventory of the system, and the accuracy and effectiveness of the proposed algorithm are verified based on simulation comparison experiments. In addition, a genetic algorithm is used to optimize the production performance of the system for a labour-constrained scenario with a reasonable allocation of the efficiency of each machine. Finally, the allocation strategy concluded from the experimental results of labour allocation provides decision support for the production operation of the system.