The gridmanager continues to run when there are no jobs to service, at least when jobs are removed with rm -forcex. The gridmanager should notice there are no more jobs and terminate.
The current behavior is intentional and happens for all grid universe jobs. We saw it as advantageous that the gridmanager continued to trying removing jobs from the remote resource after the user did a condor_rm -forcex. Some on the team disagree and think the gridmanager should give up on jobs once a -forcex is performed.