Thanks for your help.
I tried to check the issue myself - but I do not really know the background, why it does not match/match - which of the parts of the CRI are relevant and which are not. If I have more time, I'll get back and do some more reverse engineering.
The RA code seems to be compiled for JDK 1.4.
I will forward it to Progress (which is complicated :( )
Nice - this bug seems to give SonicMQ RA a performance penalty in all app servers - and they did not see it.