In the following scenario, we are seeing slowly but steadily increasing reported address usage that doesn't reset to zero when message counts go down to 0, eventually resulting in paging and other undesirable behavior.

1. Qpid / AMQP topic producer
2. Virtual topic mapping on the configured acceptor (virtualTopicConsumerWildcards=Consumer.*.%3E%3B2)
3. Openwire queue consumer matching the wildcard pattern
4. Consumption lagging production, resulting in increasing queue depth, then catching up (batch job scenario)

We can see the reported queue size / usage increase as the queue depth increases, then go back down as the queue is worked off, but not all the way back to 0. Repeating the "batch job" results in steady increase of the baseline address size (with a message count of 0), eventually resulting in paging, triggering of global-max-size limits, etc.