This change prevents the same BoundedElasticThreadPerTaskScheduler being picked up when the maximum number of Virtual Threads are already being executed in parallel. The consequence of improper busyness accounting was that tasks were executed sequentially instead of being run in parallel because the same Worker was being picked by operators.
This change prevents the same
BoundedElasticThreadPerTaskScheduler
being picked up when the maximum number of Virtual Threads are already being executed in parallel. The consequence of improper busyness accounting was that tasks were executed sequentially instead of being run in parallel because the sameWorker
was being picked by operators.Resolves #3857