Comment 8 for bug 1504725

This bug affects us as well, making unable the services such as nova* and neutron* to sustain a connection loss to rabbit twice (tried the last oslo.messaging stable/kilo from github). The rabbitmq is HA (ha queues etc), but I'm not yet sure if it's relevant.
Seems, self.channel in oslo_messaging/_drivers/impl_rabbit.py around line 865 is None after reconnecting the second time (expected to be kombu.transport.pyamqp.Channel object).
Reproducibility in my environment is 100%.