Proposing to drop the options: use_mixed_combining mixed_combining_chi_weight They haven't worked better than chi_combining for anyone yet.