The n_startup_trials' document says, "The random sampling is used instead of the TPE algorithm until the given number of trials finish in the same study.", but it is inconsistent with group=True or constant_liar=True. We need to specify the behavior for these options.