Steps for reproducing the bug:


1) Karaf commands to create reproducer env:

# start a fresh install of 6.1.0 

#create Fabric
fabric:create --profile fabric --clean --wait-for-provisioning

# install R2
version-create 1.1
patch-apply --version 1.1 "file:///path/to/jboss-fuse-6.1.0.redhat-379-r2-611423.zip"
container-upgrade 1.1 root

# install R2P6
version-create 1.2
patch-apply --version 1.2 -u admin -p admin "file:///path/to/jboss-fuse-6.1.1.redhat-442-r2p6.zip"
container-upgrade 1.2 root
version-set-default 1.2

# create broker and broker container
mq-create --version 1.2 --group TestGroup TestBroker
container-create-child --version 1.2 --profile mq-broker-TestGroup.TestBroker root TestBroker
-------------------------------------

2) further steps to reproduce the problem: 

* wait for broker container to start up fully

* run
    list -l -t 0 | grep zookeeper 
  note bundle id, assuming bundle id 53 in next commands
  
* run these two Karaf commands with max 3 sec delay between them
  osgi:stop  --force 53
  osgi:start --force 53

  Timing is crucial for these commands.  The broker container needs to print these log lines in this order:
  State change: SUSPENDED
  State change: RECONNECTED
  Disconnected from the group
  Stopping the broker TestBroker.
  Broker TestBroker is now the master, starting the broker.
  Broker TestBroker is being started.
  Broker 'TestBroker' shut down, giving up being master
  Disconnected from the group
  Stopping the broker TestBroker.

  In particular the line "State change: RECONNECTED" needs to happen before the broker starts to shut down.
  Note, the broker will be shutdown but not restarted! If it gets restarted, the problem was not reproduced. 

* If you further run
    container-connect TestBroker 
    activemq:list
  no broker is listed