I believe the root cause of this issue was the use of a single work ID for sending messages, and the fact that we would append new requests to send messages to the existing work. That meant that if one message failed to send, the rest would as well.
This should now be fixed -- we use the database ID as part of the work ID, so each message send is a unique work.
I believe the root cause of this issue was the use of a single work ID for sending messages, and the fact that we would append new requests to send messages to the existing work. That meant that if one message failed to send, the rest would as well.
This should now be fixed -- we use the database ID as part of the work ID, so each message send is a unique work.