During the setup of the Jepsen test environment, in etc/evergreen.yml, if an LXC node isn't accessible, it is restarted. The access to this node (scp) could fail if the node is up, but sshd isn't ready. A sleep or scp retry could be added to fix this problem.