In doing high-temperature/high-reliability designs, one must read and understand much more than just the data sheet. One cannot just look at the sheet and assume the part is good for, say, 125°C in one's application. Many mil-spec parts are qualified for only 1,000 hours in much of their testing over environmental conditions. This may entail a whole lot of surprises if the device must last for 5,000 hours, 10,000 hours, or even longer.
This may leave one with a whole bunch of dead parts if one does not carefully study the test and qualification data for each piece -- not just the base parameters.
In equipment used for heating, ventilation, and air conditioning, components in a rooftop unit will often make one large thermal cycle per day, as well as many smaller cycles or ripples of temperature. Design engineers must manage their own temperature tests on each component used in the product.
Typically, such testing must go through many thousands of thermal cycles to look at how aging breaks down the compounds used to make the part. These tests are intended to simulate the desired 200,000 hours or longer of thermal cycle conditions. Note that similar testing would be done for similar industrial applications, such as commercial ovens and petroleum production (well drilling).
IC molding compounds can break down. Varnishes and conformal coatings on discrete components can dry out and crack. Dendrites (metal whiskers) can form in devices. Metal migration can occur. Metal (or semiconductor material) is pushed out of its original position over time at high temperature and in the presence of sufficient electromotive force (voltage) or current flow.
Inductor and motor magnet wire will deteriorate due to fractures in the enamel or polymer coating. Rubber or plastic insulating jackets on cables or individul wires soften, crack, or dry out. In coaxial cables, the dielectric (insulation between the center conductor and the shield) deteriorates in a similar way. But with coax, this causes an additional problem: The characteristic impedance changes. This can cause additional signal attenuation and a degraded standing wave ratio with the attendant signal refections.
Rubber, plastic, cork, and other gasket materials dry out, crack, and let in water or other contaminants. Mold and fungus attack components and coatings. Humidity penetrates coatings and causes corrosion of components and solder joints. If it can go wrong, it usually does.
What does one do? Life testing -- boring and grueling life testing -- is often the only solution. For less critical designs, put less expensive electronics on a life-limited part, and replace it at service intervals. What are your experiences with long-life parts? How do you keep things running longer than normally seems possible? What deratings for parts do you consider?