hello everyone

I am investigating a problem where a remote process (spawned on a remote qnet node via "on -f") does not terminate when 
a SIGTERM is sent to on.

Here's the steps followed during a full run:
1) process procA is launched on node nodeA
2) procA spawns: on -f nodeB scriptB
3) scriptB ends into: exec procB
4) our test/simulation goes on until a good reason for terminating
5) procA sends a SIGTERM to its on child (spawned at point 2), expecting procB to get terminated
6) procA expects SIGCHLD from on, or after a timeout it exits anyway

What sometimes happens is: on remains alive with a pending SIGTERM, and procB does not terminate.

I noticed this strange message into nodeB sloginfo, in correspondence with the termination attempt:

Jan 05 17:09:42    7    15     0 qnet(stats): kif_server@899 kif_server_msgs Overflow(0)

Can this explain why something gone wrong in the signal delivery procA->on->nodeB->procB ?
Is it a non recoverable qnet error?
Is there a way to avoid this kind of error / critical situation?

I'm running a x86 system with qnx 6.5.0. If other infos are needed feel free to ask :)

thanks in advance!
Davide