On mitigating the analytical limitations of finely stratified experiments