The Platform Java SDK retains the mapping of clustername to node name in a dotfile:
$HOME\.businessobjects\clusterinfo.1400.properties
So anytime you you a node name to log on, it'll first look it up for the clustername, then pick a node at random from the mapping.
So even if node1 is down it'll try the other nodes.
If there is no clusterinfo.1400.properties file or the node1 isn't listed in there, it'll try and connect to the node1 directly.
So let's say you deploy your Platform Java SDK onto a machine where it hasn't been used before. Then there would be no clusterinfo.1400.properties file, so it needs to connect to a node to retrieve that info.
You are unable to use a clustername till that first successful connection.
By using a node name, then you're assured of connecting to that node.
Regards,
Ted Ueda