Safety-critical distributed real-time systems, such as networked medical devices, must operate according to their specification, because incorrect behaviour can have fatal consequences. A system's design and architecture influences how difficult it is to provide confidence that the system follows the specification. In this work, we summarize and discuss three design choices and the underlying concepts that aim at increasing predictability and analyzability. We investigate mandatory resource reservation to guarantee resource availability, separation of resource consumptions to better manage resource inter-dependency, and enumerative reconfiguration. We use the example of a distributed monitoring system for the human cardiovascular system to substantiate our arguments.