To find optimal strategies to control heating, ventilation and air-conditioning systems (HVAC), significant progress has been made using data-driven methods like model predictive control and (deep) reinforcement learning, which use simulation models to obtain optimal control strategies. These models simulate the system as if all components behave optimally, although in reality most HVAC systems operate in suboptimal conditions due to faults and degeneration of components. As a result, the generated control strategies are unaware of the suboptimal operation of the system. This discrepancy between the simulated behaviour and the actual behaviour of the system is called the sim-to-real gap. This paper aims to examine the impact of faults in a system on the performance of control methods that were tuned assuming the system operates optimally. A simple case study, a space heating network controlled by a PI controller, is simulated using a physics-based model. In this system, faults are introduced in the form of corrosion at the valves. The system is simulated for different levels of corrosion and different settings for the controller. Lastly, these simulations are used to assess the impact of faults on the optimal settings for the controller. They show that a setting that minimizes energy use for a system working optimally does not minimize energy use for a system that works suboptimally.