Operator training simulators are one of the cornerstones of Model-Based Systems Engineering and a key technology to enable cost-efficient, fast, safe and customizable training. Such simulators often require real-time, large-scale multiphysics models tightly integrated with Decentralized Control Systems (DCS), which poses challenges in both process modeling and communication setup. This paper identifies two major challenges: ensuring simulation efficiency by balancing model fidelity and computational cost, and achieving modeling efficiency by streamlining communication between the process model and the DCS. We highlight the complexity of managing thousands of signals and metadata, and demonstrate how communication architectures aligned with DCS templates can reduce engineering workload and support project scalability.