Function run_pool 

pub async fn run_pool(
    worker: Worker,
    count: usize,
    shutdown: Shutdown,
) -> Result<WorkerStats>

Run count copies of a worker concurrently, all watching one shutdown signal, and return their combined tally once every one has stopped.

This is queue:work --workers=N. They share a queue, not any state: the only thing keeping two of them off the same job is the driver’s reservation.