Packet of death not causing proper session termination?

[Gunther Mayer]

Hi there,

We just came across a peculiar problem whose solution is not clear to 
me: Customer logs in, gets a very long (or non-existing/infinite) 
session timeout and happily surfs away happily. Now something special 
happens and we need to terminate his session prematurely. I know I can 
do that with a packet of death sent directly to the nas on its coaport, 
BUT, I've noticed that this makes the NAS close the session only *in its 
internal state* and doesn't send an Accounting-Stop packet back to 
radius to update session parameters (most noteably input&output octets). 
I know I can close the session *on the database* manually but I 
obviously lose those up-to-date session parameters.

So, how can solve this dilemna? It would be nice to be able to force 
chilli to terminate the session properly upon receiving a packet of death...

Gunther