Problem

SeverCounters.TotalOutstanding was not including messages in the PhysicalBridge backlog queue. The PendingUnsentItems property was never being set, so it always returned 0.

This caused a significant issue when using TotalOutstanding for connection load balancing in our PROD runtime: a connection that was stuck or slow (with messages accumulating in the backlog) would still report TotalOutstanding ≈ 0, making it appear healthy and causing more traffic to be preferentially routed to the unhealthy connection.

Fix

Propagate the _backlogCurrentEnqueued in the ConnectionCounters returned by PhysicalBridge so that the number of messages in backlog queue can be exposed to end user