For instance: have a loop of two push requests (via java sender), sending the same payload. Internally the endpoint issues to different 'submit' to the router, but than, the Message Queue is bound to the variantID. Of course, this is the same for different push jobs.

RESULT:
in case of the loop of two, the data for the first request is 2 btaches and 0 for the second.
If you do a sleep(1000) between the sending of the requests, you get 1 batch for each (as expected). I think... we need to have something like variantID+pushjobID to make this more robust