Running while ./perftest --count 100000 --nsubs 4 --npubs 4 --summary; do true; done against a clustered broker should keep going, but fails after several runs.
How does it fail? Any log file entries? --log-enable=debug+:cluster --log-enable=info+ gives better information on what the cluster is doing without imposing a per-message logging overhead.
Various framing related errors implying some sort of corruption of buffers before decoding, e.g.: 2008-dec-03 06:14:33 critical 20.0.10.15:21718(READY) error in cluster delivery: framing-error: Frame size too small 0 (qpid/framing/AMQFrame.cpp:88)
This is related to, and possibly caused by an aisexec issue: https://bugzilla.redhat.com/show_bug.cgi?id=474277
*** Bug 472242 has been marked as a duplicate of this bug. ***
Fixed in rev 732768 Added flow control options with default values that work reliably.