What you're seeing is the expected behavior in certain circumstances [1].
In your case, you're sending a durable message to the broker and the
connection fails. The problem now is that the client doesn't know what
happened to the message. Did it make it to the broker and get persisted to
disk befo
Yes, I too faced similar issues while randomly killing artemis instances.
The pattern which I observed was mainly, when you kill live instances in a
cluster w/o running backup node. So, in our case we incorporated *client
side failover* detection. IMO, most of the times; failover happens
transparen
