While we already have a flag in place to limit how many instances of a task (and its parameters) are allowed to be queued we also need a flag that limits how many instances of a task can be worked on concurrently at the same time (for example by a fullctl_poll_tasks job running multiple workers).