<hardware> It is common lore among hackers (and in the electronics industry at large) that the chances of sudden hardware failure drop off exponentially with a machine's time since first use (that is, until the relatively distant time at which enough mechanical wear in I/O devices and thermal-cycling stress in components has accumulated for the machine to start going senile). Up to half of all chip and wire failures happen within a new system's first few weeks; such failures are often referred to as "infant mortality" problems (or, occasionally, as "sudden infant death syndrome").
See bathtub curve, burn-in period.
Try this search on Wikipedia, OneLook, Google
Nearby terms: Industry Standard Architecture « inetd « inews « infant mortality » infeasible path » inference » inference engine