BTW, the thing that often kills ICs/LEDs/etc. is excessive heat. This occurs when the product of the current through a part of the device times the voltage across that part exceeds a reasonable value. Note that it's also not a simple relationship between power and failure. For example, a high power level can result in a conductor in the device dissipating too much power and catastrophically failing (e.g., through melting). A lower power level won't cause this melting, but can result in a different failure mode (say, something like electromigration). I haven't worked on IC reliability for a number of decades, but even back when I was, there were many different failure modes, many of which were thermally activated (i.e., there was an Arrhenius exponential relationship for the time to failure as a function of temperature).