So the primary was much beefier than the secondary? If so, why let it fail over? I figured you should always have the same resources for a mongo primary and secondary.