Currently, the attributes for each child resource of a stack or channel use RESTART_ALL_SERVICES, thus any change to one of these attributes requires all services to restart. This is overkill, since only the services of the parent resource (i.e. channel, channel factory) actually require restarting.