Usually when something like this happens to us it is because
we have resent a smat message through the thread.
The tps procs generate an ACK and that is the message that is
hanging out there. Since the sending system on that thread isn’t
expecting an ack, it just stays there.
Sometimes the sending system will get screwed up with this
extra ACK and hang/pause/or shut down.
Since “use recovery database” does not apply to these messages,
we can simply bounce the interface and they’re gone.
