Like others have reported, I have seen failures with this test after upgrading to the recent RE_24. I have only seen it when running the test on a multi-cpu linux system tough, it works find on a single-cpu solaris.
When it fails it is always syncrepl in the two first servers that fails to bind to the third. In the logs they complains repeatedly like:
slap_client_connect: URI=ldap://localhost:9013/ DN="cn=config" ldap_sasl_bind_s failed (-5)
I.e, the bind timed out. Server3 logs the bind request, but first after the unbind that follows.. Modifying the source with some more debug output that prints the time it shows that the bind request is sent and times out (after 3 seconds) before server3 logs accept of the connection. Some race condition that kicks in if two servers tries to connect and bind at the same time? Btw, I've only tested with back-bdb.
Well, that's how far I've had time to look at it. I have plenty of logs available if anyone care to look at them.
Rein