Queues should raise 'event' to endpoint when 'message' added
Polling is a clunky alternative (unnecessary charges, bandwidth consumption etc for queues with little indeterminate levels of activity) and if the endpoint remained internal to say a load balanced worker role, the platform might be able to liaise between the queue service and idle/new worker instances automagically.
We want to build a solution where our SLA's are message driven which roughly equates to a queue per message type. We'd like to keep queue to worker time twithin an acceptable boundary and be able to fire up new workers when the time in a queue exceeds this level. It would be very helpful it we were able to monitor the average time a message lives in a queue in near real time (or time based triggers would be even better) so we can spin up new nodes when needed.