Comment 3 for bug 183568

The problem appears to be caused when network manager catches the SIGTERM (15) signal
I have a workaround involving hard killing network manager with SIGKILL:

Modify /etc/dbus-1/event.d/25NetworkManager as follows (just add the --signal 9 option)

d_stop() {
        # Modified MF 11 Feb 2008 To avoid NM hang
        #start-stop-daemon --stop --retry 60 --quiet --pidfile $PIDFILE \
        # --oknodo --user $USER --exec $DAEMON

        start-stop-daemon --stop --signal 9 --retry 60 --quiet --pidfile $PIDFILE \
                 --oknodo --user $USER --exec $DAEMON
}

Modify /etc/init.d/sendsigs as follows
...
        # Kill all processes.
        log_action_begin_msg "Terminating all remaining processes"

        # Added MF 11 Feb 2008 To avoid shutdown hang
        OMITNM=
        if [ -f /var/run/NetworkManager/NetworkManager.pid ]; then
                OMITNM=$(cat /var/run/NetworkManager/NetworkManager.pid)
                OMITNM="-o $OMITNM"
        fi
        echo "Not killing $OMITNM"

        killall5 -15 $OMITPIDS $OMITNM

        # END MF mods
        log_action_end_msg 0
...