The current behaviour is described here: https://issues.opennms.org/browse/NMS-10637
In notifications you may want to have the IP address of a node. Which can be very difficult depending on the node.
In case of server nodes which often have only 1 IP interface, it's not hard to get this information. But what if you're dealing with switches which have many interfaces?
I can understand why interfaceDown or nodeLostService events are able to translate %interface% and %interfaceresolve% since there is a direct relation between the outage and the interface.
But it can look very different when nodeDown occurs (as I said, depending on the node).
My suggestion is to fill the variable with the primary managed IP interface of a node. This would fit to all kind of nodes.