First of all, sorry to take so long in reply to this question. And it is unfortunate to see that the information in your screenshot is really confusing when the overall health shows "Green" and the instance in the listing shows "Offline"
I don't think you did anything wrong here. I think it's a problem of the method that the code uses to try to "clusterize" the Kibana instances reporting data. We tried to follow the look and feel that Elasticsearch has of showing cluster health and node availability, but the fact is that Elasticsearch is a much different product since the connected nodes are clustered in the true sense of the word, but Kibana instances are not.
We'll have to do some re-thinking and also gather more data about each Kibana instance to give a better view of the historical state of Kibana instances both individually and as a group.