Interesting. Has anyone found running multiple parallel agents useful in practice?

You'd save time compared with running them in serial obviously?