Instead of the FE to drive the behavior of On-Line Backends, Implement a Polling model where the FE queues requests for the BE and the BE fetches new requests as it process them. This allows to better align (for what is possible) the offline and the online models.
Instead of the FE to drive the behavior of On-Line Backends, Implement a Polling model where the FE queues requests for the BE and the BE fetches new requests as it process them. This allows to better align (for what is possible) the offline and the online models.