The deployment of configuration and the restarting of services on large swift clusters must be carefully orchestrated and managed by a swift knowledgeable operator.
There should be a "production" flag that disables all swift-related service management in chef for those large deployments.
This includes syslog-ng as well as the general swift services, but not necessarily supplemental services like keystone, mysql, ntp.
This should be part of the "scalable swift" milestone.