I have pretty much exactly one album that really has a peak to average ratio high enough to introduce clipping when ReplayGain is used (Vangelis: Opera Sauvage, album gain +4.69 dB but peak @-1.20dBFS). Otherwise album gains of 0..-3 dB are typical for old productions.
My point with ReplayGain is the following: If it's commonly used, volume on loud albums will be brought down during playback anyway, and that in turn makes overly hot mastering redundant, leaving only the negative effects (the clipping) without any gains. In such a scenario it seems much smarter to leave the kind of headroom that is needed.
Originally Posted by rederanged
So it sounds like it can be of some help if the clipping is induced by lossy compression, but it's not going to do much for you if the original recording is clipped, which is the case for a lot of modern CDs.
Yep. Actually the peak value in an MP3 of a certain VBR quality level when compared to the same in the original track can serve as an indicator of how heavy the brickwalling was. One track in the aforementioned Cocteau Twins album has a ratio of about 1.35 at -V 4 -q 0, while a more moderately brickwalled album might have like 1.15...1.25 (it doesn't say anything about clipping already introduced during recording though). Very echo-y '80s albums tend to be a bit under 1.