Most system resolvers do not keep track of resolver performance across queries. If, for example, the first resolver listed is not functioning, many system resolvers will still try it first on every single resolve, moving on to another resolver only when that server times out -- over and over. This reduces the time available to complete a resolve, increasing the chances that hosts that are slow to resolve won't resolve at all.
ConferenceRoom's internal resolver can easily query more than three nameservers, again unlike most system resolvers. A performance history of each name server is kept, so that queries can be sent to the nameserver that is responding fastest. Load can be distributed if the number of resolves requires is great enough to affect performance of the name server.
This resolver is especially important for the chat server. When a busy server is restarted or when a network outage is repaired, the limiting factor to how quickly a server can take clients back on is often how quickly it can resolve them.