Currently, GnuTLS handshakes are synchronous (blocking) in `lrmd_tcp_connect_cb()`. If two remote connection resources are started at roughly the same time on the same host, the first could block long enough to cause the second to fail. The fix will be to either make the handshake asynchronous (a more comprehensive solution) or possibly to serialize remote connection starts (possibly an easier solution).
See also:
* T855
* [[https://issues.redhat.com/browse/RHEL-34276 | RHEL-34276]]