stan returns metric it will use, after warmup period, with which it would sample during sampling period. ideas like metric sharing can shorten the estimation time -- but without much success during the last experiment.
{"inv_metric": [0.00890483, 0.0146411, 0.0180988, 0.0173486, 0.0162696, 0.0185366, 0.0186761, 0.0176091, 0.0177719, 0.0194737, 0.0269911, 0.0241012, 0.0270683, 0.0305894, 0.0295884, 0.0285179, 0.0270491, 0.0247198, 0.00094971]}