Recent industrial applications are implemented in a modular way, resulting in flexibility during the whole life cycle, i.e., setup, operation, and maintenance. This applies especially to larger applications like logistic, production, and printing processes. Their modular character is resulting from the constantly increasing complexity of such installations, which makes their supervision for securing reliable operation a difficult task: the data of hundreds (if not thousands) of signal sources must be acquired, communicated, and evaluated for system diagnosis. In this contribution we summarize the challenges arising in such applications and show that distributed sensor and information fusion for modular self-diagnosis tackles these challenges. Here, we propose an innovative distributed architecture encompassing intelligent sensor nodes, self-configuring real-time communication networks, and a suitable sensor and information fusion system for condition monitoring. New challenges arise in the context of distributed information fusion systems, which are identified and to which an outlook on future solutions is provided. A number of these solutions have already been discovered, implemented, and are evaluated in the context of a demonstrator, which resembles a real-world printing application.