Triggers and Recursion

[James Lampert <jamesl@xxxxxxxxxxxxxxxxx>]

We've had a recent report (naturally, while I was on vacation) of this 
exception occurring with an "AFTER" trigger (the names have been changed 
to protect the innocent):

> Message ID . . . . . . : CEE9901 Date sent  . . . . . . : 09/25/08 Time sent  . . . . . . : 12:13:45                                                                             
> Message . . . . :   Application error.  RNX8888 unmonitored by FOOBAR at
> Statement *N, instruction X'0000'.


> Message ID . . . . . . : RNX8888 Date sent  . . . . . . : 09/25/08 Time sent  . . . . . . : 12:13:45                                                                             
> Message . . . . :   FOOBAR was called recursively.                         

The trigger program (called FOOBAR here) is mixed-language, with an ILE 
RPG main calling ILE C functions and procedures.

Now the normal (and desired) behavior here is for recursive triggering 
to just fail quietly, rather than lock up the job, or (worse) go into a 
potentially system-crashing recursion loop. So far as I'm aware, this is 
an isolated incident, but I vaguely recall hearing of similar incidents 
in the past.

I quite honestly have no clue what could be going on here. Can somebody 
point me in the right direction?