Running 2.3.31, I'm seeing the behaviour described in ITS4790, in which Pierangelo Masarati wrote:

...

Apparently, one thread is unable to terminate correctly... but then slapd hangs forever in ldap_pvt_thread_pool_destroy().

and Howard responded:

This is a bad patch to connections_shutdown in RE23, there is a "continue" after checking for a client session that shouldn't be there. This line is not present in HEAD and the bug doesn't exist in HEAD. Strange that it got there since we just recently sync'd RE23 with HEAD's connection.c.

Ok, so the problem line is in connection.c, right?

I know just about enough C to be able to comment out the line. How do I identify it?

I need to fix this pronto, as I have already rolled out the bad code to about half of our ~80 replicas.

Cheers, Lesley W (I can just hear "I told you so..." coming from my employer when he hears about this)