Developments in communication and automation have driven the expansion of distributed networks, essential for IoT and CPS development in industrial applications requiring reliable image processing and real-time adaptability. Although broadly adopted, there is a literature gap regarding the performance of MQTT protocol for image sharing and transmission under high-traffic scenarios with intermittent connectivity, restricting its use in critical IoT and CPS applications. In this context, the present work examines the reliability of real-time image transmission in IoT and CPS industrial systems that utilize the MQTT-based publish/subscribe communication model. It focuses on scenarios with network interruptions and high data traffic, evaluating the performance of a distributed system through a series of controlled testbed validation experiments. Experimental validation demonstrated that while the MQTT-based system sustains reliable transmission under normal conditions, its recovery capability depends on the failure point, with complete restoration occurring when disruptions affect the Orchestrator Node and partial recovery when the Producer Node or Broker are affected. The study also confirmed that the system prevents duplicate errors and adapts well to increasing network demands, reinforcing its suitability for industrial applications that require efficient and resilient data handling.