A regression from https://issues.apache.org/jira/browse/AMQ-5794 .
Connection watchdog is started for every initiated connection and stopped on WireFormatInfo command. HTTP transport doesn't send WireFormatInfo so the watchdog never realises that the connection has been successfully established.
The connection gets terminated every 30seconds by the watchdog.
At the beginning, everything looks fine, but then you start getting exceptions and start losing packets. I haven't seen that myself, but I had people reporting that if HTTP transports are in use, it eventually destabilises the broker and affects non-HTTP transports too.
2016-04-22 10:32:46.029244500 2016-04-22 10:32:46,029 WARN [ActiveMQ InactivityMonitor Worker] [Transport] Transport Connection to: blockingQueue_28120594 failed: org.apache.activemq.transport.InactivityIOException: Channel was inactive
2016-04-22 10:33:30.988313500 org.apache.activemq.transport.InactivityIOException: Channel was inactive (no connection attempt made) for too (>30000) long: blockingQueue_21644517
2016-04-22 10:33:30.988637500 at org.apache.activemq.transport.AbstractInactivityMonitor$1$1.run(AbstractInactivityMonitor.java:91)
2016-04-22 10:33:30.988667500 at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
2016-04-22 10:33:30.988689500 at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
2016-04-22 10:33:30.988717500 at java.lang.Thread.run(Thread.java:745)