We currently discard messages in a container if it is not fully started. In a master-slave scenario, either master or slave will discard messages sent by the other who starts first. The current situation is that slaves are inited and then master is marked as running. This causes very early messages in slave container (if it starts very quickly) to be discarded. This PR fixes the situation by allowing the master container to deliver the message to the agent, and having the agent queue it until it is ready to respond.
We currently discard messages in a container if it is not fully started. In a master-slave scenario, either master or slave will discard messages sent by the other who starts first. The current situation is that slaves are inited and then master is marked as running. This causes very early messages in slave container (if it starts very quickly) to be discarded. This PR fixes the situation by allowing the master container to deliver the message to the agent, and having the agent queue it until it is ready to respond.