In this paper, we consider opportunistic downlink scheduling in a cellular network that exploits multiuser diversity using one-bit feedback. To reduce the feedback overhead inherent in opportunistic scheduling, mobile stations are allowed to send one-bit information to the base station only when their channel quality exceeds a given threshold value. The objective of this paper is two-fold: (i) find the threshold value that can optimize queueing performance of the scheduler and (ii) investigate the relationship among capacity, fairness, and queueing performance. To this end, we first derive a formula for the sum-rate capacity as a function of the threshold. Next, we quantify the long-term and short-term fairness using JaM's fairness index. Lastly, we develop a packet-level queueing model and derive QoS measures such as queueing delay and packet loss probability. Based on our analysis, we optimize the threshold value by considering both the traffic condition at the MAC layer and the temporal channel correlation at the PHY layer. Numerical results show that the optimized threshold can significantly reduce the average queueing delay and the packet loss probability. In addition, we find that there is a trade-off between short-term fairness and sum-rate capacity, whose control knob is the threshold. We show that, from a queueing performance perspective, supporting the sum-rate capacity (resp. the short-term fairness) is more important when the traffic is heavy (resp. light) or the wireless channel varies with low (resp. high) temporal correlation.