Scheduling real-time Quality of Service (QoS) data flows in a large-scale wireless communications system uses credit allocation to active QoS flows based on quality-of-service weights, and the credits are measured in dimensionless units of air interface slot capacity. Scheduling addresses inactive QoS flows with no data pending for transmission that may accumulate up to a burst credits limit based upon a fair share of all credits that would be allocated in a 100% busy system. Intermittent flows such as echo requests or keyboard input can thereby get immediate service by using their burst credits. The flow with the highest credit accumulation is serviced first and may use up to a full time slot (or system quanta) of air interface time reducing header encapsulation overhead when payload sizes (and spectral efficiency) are low. Additive and multiplicative aspects age the negative credits back to zero whenever a flow overspends its credit allocation.