System test plays a very important role in the product development cycle of a safety-critical device, such as an insulin pump. Given the significant risk of this device, fault-injection should be performed to validate dependable implementation under abnormal circumstances, namely: data corruption, bit flips, incorrect signal assertions and more. Clearly, a robust testbed is essential to create the required fault conditions at the system level. In this paper, we consider a Hardware-In-Loop (HIL) testbed to perform system testing of a prototype insulin pump. We compare different hardware and architecture possibilities for implementing the testbed and conclude that a modular architecture (MAESTRO), that combines event-driven and time-driven operations is most optimal. This testbed is implemented on an FPGA, providing better performance and feature sets, especially if systems with multiple microcontrollers are targeted. In particular, MAESTRO allows us to inject faults based on events, as opposed to faults scheduled for specific instants in time, which provided resilience to task jitter and software modifications.