"In your scenario, we have already excluded the possibility that the parts themselves are bad, at least according to the way we measure 'bad' or 'good' in-house. Before I start chasing ghosts, I want to make sure that the customer is using the part correctly. Was there an undocumented change in the service environment such as a nearby leak or excessive temperature? Did they follow the wiring diagram? I would provide a few brand-new parts that we had verified were 'good,' and ask the customer if I could tag along and watch as they were installed and tested, so that I could make sure that we had captured all the requirements properly. If I could see something obvious during that phase of the problem-solving, I would bring it to their attention, and we would discuss whether the customer's written requirements needed to be updated, leading to a change in the part, or whether they could adapt to using the part as intended. If everything about the way they used the part was in order, I would want to document every detail of their real service environment and take it back to our product team for them to study. Is it possible that our in-house evaluations did not test the characteristic that was determining success or failure in the service environment? That's a possibility that has to be considered. These are the first few steps to how I would attack the problem you outlined."