SAA/RTR/IPSLA - service unavailable traps?

I've been experimenting lately with Service Assurance Agent (aka RTR,
aka IP SLA, depending on IOS version). So far I have it monitoring a
few things happily, and I have Cacti graphing the results of those
tests, but I haven't had much luck getting the 'reaction' part of the
system going.

What I'd like is to receive a trap from the router when the service is
unavailable, as well as falling outside of parameters. In particular,
for icmpEcho, http and tcpConnect tests, it seems that the far end
being unavailable *doesn't* send a trap, although it does show as an
error in 'show rtr op'.

The Cisco docs aren't exactly clear on whether it is supposed to do
anything in this situation, but it seems that "no service" is the
worst SLA-failure of all, to me Has anyone got a sample config for
this that does work?

I have:

snmp-server enable traps rtr
rtr logging traps

! my test webserver
rtr 13
type tcpConnect dest-ipaddr 10.1.1.188 dest-port 80
tag mac2
rtr reaction-configuration 13 react connectionLoss threshold-type
immediate action-type trapOnly
rtr reaction-configuration 13 react timeout threshold-type immediate
action-type trapOnly
rtr schedule 13 life forever start-time now

but regardless of whether I shutdown just the http service (so the
probe should get a connection refused), or pull the network cable (so
it gets nothing), I don't get a trap at my NMS. Traps generally are
working fine for my test router though, so I know it's not that. Test
box is a spare 1751 running 12.3(11)T, but I haven't had any joy with
12.4 either.

Anyone else do this successfully?

Howie