Comment 4 for bug 1676464

This is due to a bug in the startup code. The two slave threads are waiting on a condition variable for wsrep_ready. However, when wsrep_ready is changed, it uses mysql_cond_signal() which wakes up a SINGLE thread. So one of the slave threads is signalled and continues, the other thread is waiting on the condition variable, which will not be signalled again.

The solution is to use mysql_cond_broadcast() to wake up all threads waiting on the wsrep_ready condition variable.