There is an issue where people will begin having issues accessing Sametime if one server in a Sametime HA environment is down beyond a certain length of time when the Server Community DNS is set up as round robin.
Essentially either the clients or the servers “forget” about the downed server. You are OK for up to some number of days where they recognize that the server is a ST server but no longer available. After that some cache is cleared and this problem arises.
This is why normal HA works just fine (when a server goes down for just a few minutes or hours) the servers and / or the clients are able to ignore the proffered IP address since they are aware that the server is not longer responding.
If you are taking a server out of service for longer than a week (or the period within which you will be recycling all of the other servers in the community), remove the downed server from the round robin DNS until you are ready to put it back into service.