I guess one trick would be to size the compute environment so that it any has enough compute to run one job at a time, and then batch won’t start a second