In this paper, we consider an unreliable production process which produces nondefective items when operating in control, but produces defective items with a probability α when the process has shifted to an out-of-control state. Following a JIT philosophy, we stop the entire line and repair the machine as soon as we detect that the process has shifted to an out-of-control state. To test whether a process shift has occurred, we inspect the last m units for every n units produced and stop the machine if a defective unit is found. More important, we place a ‘delay buffer’ immediately after the unreliable process, which serves to delay the movement of items from the unreliable machine to other processes (or customers) downstream in the production system. When we detect that the machine has shifted to an out-of-control state, we stop the entire line and examine all previously uninspected items in the delay buffer; in this way, the buffer serves to reduce the expected rework and penalty (e.g., warranty) costs downstream when a process shift has occurred. In this paper, we develop a model for this approach and use this model to test the operating characteristics of our system. Computational results illustrate our hypothesis that a delay buffer may significantly reduce expected total costs of a quality control process.