When the TransferQueueBundler ("new" / default) sends bundled messages, if there is only a single message in the list, the message is still sent as a batch (of 1).
Currently, parsing of single messages is done in a separate thread (JGRP-1716) whereas parsing of message batches is done in the thread which received the messages from the socket.
If we therefore send a batch of 1 as a single message, we might get better perf (to be verified).