The only way to know would be to look inside. If you want and are willing to ship the dead chips to me in the US/Silicon Valley, I could put them in the X-Ray machine and see if there are any obvious bond wire faults. But the wires might still look OK in the X-Ray micrograph if the defect is due to development of an oxide film between the pad of the circuit and the bond, or a chemical diffusion or reaction in the bond pad itself. There are (very expensive) tools that can ferret out such material problems (after you crack the chip open), but I am not sure I can get easy access to them. Maybe I can ask someone at National Semi (now TI) down the road from us and see if they'd be interested to research the cause of failure. Would be cool to do an electron micrograph section and find out...