Actually this is probably more complicated since it has to account for rescheduling or not, so it might require handling teardown like some of the rescheduling code in the _do_build_and_run_instance method in the compute manager.
Actually this is probably more complicated since it has to account for rescheduling or not, so it might require handling teardown like some of the rescheduling code in the _do_build_ and_run_ instance method in the compute manager.