If A sends 100 multicast messages (or unicast messages to self), the messages to self are not batched but looped back individually, each using a separate thread from the thread pool.

So in the above case, A will receive its own 100 messages using 100 threads from the pool.

It would be more efficient to batch the 100 messages into a batch and then loop the batch back, taking up 1 thread instead of 100!

To do this, looping back has to be done after the bundler, when the bundler is sending the message batch.