having had a quick look at MP3gain and the replaygain algorithm the default is normalising to somewhere around -14dBFS. Which (without getting into the heavy maths I barely understand myself) and converting roughly into loudness units (-14 LKFS) is just right for “youth orientated music stations”, according to the Scandinavian public broadcasters advising the EBU.

Originally they suggested -23 LU for all stations but decided to let them go a bit louder (the full details are in the papers sinner posted a while back.

I’ve been monitoring the current live broadcasts and found that those using vinyl gramophone records source material tend to settle around -23 LU and those using digital sources are at -14 LU. To my ears either sounds perfectly acceptable and -23LU gives headroom for vinyl surface noise. Younger people who have grown up with or prefer genres mostly released as digital are going to prefer -14 LU which is about the loudest without sacrificing the source dynamic range by using compression.

my view is a file which is triggering the clipping detect is already biting into the headroom and best left as it is. Otherwise you are pushing it to the point it goes beyond 0dBFS which is actually possible but how soundcards interpret this is undefined (and could be unpleasant to people ears and their loudspeakers!). We could try one and host it somewhere with suitable warnings attached, but the Northern Europeans view is “don’t do it”.