Early main-frame system administrators had the job of taking program execution requests, assigning them a priority, and giving them a time frame to run.
The original implementation of LLQ over Frame Relay on these platforms did not allow the priority classes to exceed the configured rate during periods of non-congestion.