Problem: many regulus operations still assume KUBECONFIG already set in the .bashrc, and therefore they invoke "ssh user@bastion 'kubectl ....'" - This will not work when the bastion hosts more than one clusters where KUBECONFIG must be
set per login shell and not hard set in the .bashrc.
To do: improve those deficienced ssh'es. Make them read lab.config for KUBECONFIG. Use this form, ssh user@bastion "cd $reg_dir && source bootstrap.sh && kubectl ... ", already exists in many other places.