Motivated by the Internet of Things (IoT) and Cyber-Physical Systems (CPS), we consider dynamic wireless fading networks, where each incoming flow has a random service demand and leaves the system once its service request is completed. In such networks, one of the primary goals of network algorithm design is to achieve short-term fairness that characterizes how often each flow is served, in addition to the more traditional goals such as throughput-optimality and delay-insensitivity to the flow size distribution. In wireline networks, all of these desired properties can be achieved by the round-robin scheduling algorithm. In the context of wireless networks, a natural extension of round-robin scheduling has been developed in the last few years through the use of a counter called the Time-Since-Last-Service (TSLS) that keeps track of the time that passed since the last service time of each flow. However, the performance of this round-robin-like algorithm has been primarily studied in the context of persistent flows that continuously inject packets into the network and do not ever leave the network. The analysis of dynamic flow arrivals and departures is challenging since each individual flow experiences independent wireless fading and thus, flows cannot be served in a strict round-robin manner. In this paper, we overcome this difficulty by exploring the intricate dynamics of TSLS-based algorithm and show that flows are provided round-robin-like service with a very high probability. Consequently, we then show that our algorithm can achieve throughput-optimality. Moreover, through simulations, we demonstrate that the proposed TSLS-based algorithm also exhibits desired properties such as delay-insensitivity and excellent short-term fairness performance in the presence of dynamic flows over wireless fading channels.