Thank you very much Jens for your detailed and comprehensive bug report! 
I really appreciate your feedback.
The intentation to use a non-single threaded ExecutorService here was to gain some speedup from parallelizing the PacketListener invocation. But I see now that it comes with the unwanted side-effect that the listeners could be invoked out of order.
Logged as 
SMACK-583 and commited your suggested fix.