Sometime CAFE will fail with highly variable orthogroups. So the best option as per their docs is to trial removing just the top OGs that cause the code to fail.
So we need to add this functionality, that it will try with all the data first, then if it fails due to this issue, to retry with those offending OGs removed. COuld start with say top 10, then try againe if that fail try removing top 50... Something like that