I'm having a rather strange issue when trying to migrate a resource
between members of a 3-node cluster. From all appearances the clustered
volume and its IP address load cleanly on all three node members. The
resource is reachable when it is loaded on either of 2 nodes in the 3
node cluster. On the node that it doesn't work, the IP address is
loaded, and pingable by all cluster nodes, but isn't pingable from
outside the subnet. If I migrate the resource back to either of the two
working nodes, it is pingable and I can mount the volume from a client.

Any thoughts about what might be happening here? This is a three node
cluster running NW6.5 SP3, with all applicable post SP3 patches applied.