From Customer
Interesting point here is, current_workers_count is ~10X compared to active_workers_count; means no tasks assigned to 10X threads.
How did threads land with no tasks? Is there orphaned threads not getting cleaned up?
My Notes
The total workers are divided evenly among the visible schedulers (schedulers that accept incoming query requests.) For example, a max worker setting of 1000 on a 4-scheduler system is 250 per scheduler. Each scheduler maintains a minimum level of workers, so workers don’t have to be created when new work arrives. You can think of them as hot workers.