AFAIK, I have never had to reboot the cluster when adding / deleting NTP server on a server in the cluster. I would be interested to see if there's any documentation mentioning the same.
You can SSH to each server and run 'utils ntp status'. The * beside the NTP server is the one the server is trying to sync to. The output also shows if the server in synchronized / unsynchronized with the NTP server.
By architecture, the Pub is the only server that can be configured with an external NTP. Any sub in the cluster will sync to the Pub's time, by design. This is to ensure that all servers have the same time, for replication to work fine on the cluster.
The defect Rob has mentioned in pretty severe - I would suggest upgrading the servers to a later version (6.1.5 preferrably) if possible, to avoid running into known, severe defects.
Also, keep in mind that Windows NTP servers are not supported with Linux CCM servers