Comment 17 for bug 1707160

The issue is extensively being discussed in #openstack-oslo; logs: http://eavesdrop.openstack.org/irclogs/%23openstack-oslo/%23openstack-oslo.2017-08-02.log.html#t2017-08-02T19:18:05

Current thinking is that we hit a race between

1) rabbitmq cleaning up the topic after rabbit_transient_queues_ttl [30 mins] since the last usage by a neutron agent (new agent doesn't use it anymore) and
2) oslo.messaging sending a fanout to a topic.

In regular setup, a fanout to a non-existent topic results in its creation; but if it is pulled from under fanout attempt in the middle of it, then we may hit the issue.

A possible solution would be then to catch the error, do any necessary cleanup for connection channel, and repeat the attempt, now first creating the topic.