HA rgmanager continuously stops and re-starts brokers.

On a test cluster of 3 nodes, rgmanager continuously kills the primary broker shortly after starting it. The cluster never settles down in a stable configuration. See attached cluster.conf

This does not seem to be a regression in the HA code the same problem occurs with qpid code going back to June 17 2013. It may be a change in behaviour of rgmanager, or possibly an unintended change to the cluster.conf (attached.)