The use of intentional time delays is considered in continuous-time feedback control systems. It is shown that ARMA controllers have little advantage over simpler MA controllers, so long as no limitation is imposed on the number of available time delays. A necessary and sufficient condition is derived for a closed-loop system with an ARMA controller to be practically stable. This leads to a drastic restriction on the use of ARMA controllers, i.e. an ARMA controller must be practically stable itself. It is also shown that a practically stable ARMA controller can be replaced by an MA controller while preserving practical stability and giving an arbitrarily small change in the shape of the closed-loop sensitivity.