Look into including the replicated SFSB cache's buddy replication group topology in clustered SFSB proxies, and using that information in fail over decisions.

This is basically an optimization such that when a failover occurs, data gravitation can occur in VM rather than over the network.