Recent shop floor paradigms and approaches increasingly advocate the use of distributed systems and architectures. Plug-ability, Fault Tolerance, Robustness and Preparedness are characteristics believed to emerge by instantiation of these fundamentally new design approaches. However these features, when effectively present, often come at the cost of a greater system complexity. Enclosed in this complexity increase is a plethora on unforeseen interactions between the entities (modules) that compose the system. The purpose of this paper is, in this context, twofold: to validate a fault propagation model in random networks (that simulate the connectivity of modular shop floor systems) and assess the performance of two diagnostic approaches to expose the impact of relying in local or global information.