This setting controls the maximum size of the mapper pool queue in each worker. The mapper pool's size is by default set to 50% of the cores on the node. This size can be overridden by QUERY_EXECUTOR_CORES. The queue is configured by default such that the mapper pool will have 1 queue slot per pool thread. This configuration ensures that mapper pool threads do not run out of work before the scheduler main loop can enqueue more.

It is recommended to leave the setting at the default value. Decreasing the queue size runs the risk of leaving the mapper pool intermittently idle while the scheduler prepares more work to execute. Increasing the queue size increases the latency on query prioritization decisions, since it takes longer between the scheduler deciding to execute a piece of work, and that work making it through the queue.