Reliability analysis is a design-time activity. The goal is to anticipate problems and suggest product improvements before manufacturing and delivery. Enhanced quality and customer satisfaction along with reduced life-cycle costs are the motivation. Analysts employ models of physical systems to determine points of vulnerability and failure probabilities. Unfortunately, the models of complex systems--those with dependency cycles among components--necessarily lead to ambiguous analyses. Below, the source of these ambiguities is identified and some AI and expert systems techniques are suggested to help deal with them.