This is by design and is used as a node isolation mechanism.

When the node has no access to second node (ring) and ping nodes it will reboot in loop until the ring returns. We have to use reboot mechanism because our kernel does not support shutdown.

Attempt to join the node to the cluster is performed only after first reboot. If node after first reboot still does not have access via ring, second reboot (with ring returns) will not perform join cluster action.