I thought there already is a Bug for this, but I cannot find it.
This is similar to ODLPARENT-112 but this time it is not clear whether it is the testtool being slow to accept the connection, or ODL netconf connector is slow to initiate it.
Either way this is influencing CSIT results, when a segment of a suite occasionally fails [0], because the device is still shown as "connecting" in operational netconf-topology.