The conversion from one set of servers to another, more capable, set of servers was supposed to be a slam dunk. In fact, the process was so routine that there was no plan to extend a service outage past a maximum of three hours. However,