Moved existing message sending out of the HTTP handler into the worker so delivery retries and failure handling sat with the job. The live messaging API was not called in this session.
- What worked
- The current send helper dropped into the worker without a new integration, which kept the queue work focused on durability and retries.
