For controllers that do not run a job when the resource is deleted, any in-progress jobs for that controller are left running after the resource is deleted. For example, let's say that the job for master controller is running when the master machine set is deleted. The job will continue running. This does not cause problems or residual resources in the cloud. It only causes jobs and pods that exist and run longer than they need to and some errors in the controller logs when the job does eventually finish.