I would listen to whatever setting sounds better. It's hard to predict what that will be, especially without knowing what circuit gymnastics are involved on making the 4/16 switch.
If you want to baby the amp above all, try measuring the heat sink temp after extended playing at each setting. I believe this uses a simple bias scheme not unlike Krell's sustained plateau bias, so it will run somewhat hotter at high volume by design.
If you want to baby the amp above all, try measuring the heat sink temp after extended playing at each setting. I believe this uses a simple bias scheme not unlike Krell's sustained plateau bias, so it will run somewhat hotter at high volume by design.