Any test involving a change in loudness must be level matched before we try to figure out anything else by ear!
Otherwise we naturally get all sorts of impressions, but they're caused, or at least strongly affected by the difference in loudness. The forum is full of false testimonies and misinterpretations around replaygain, gain switches, balanced VS single ended output, and even plain old volume pot.
Replaygain by default will add a value in the metadata to tell the player app how to adjust the volume if that tag is requested. a track will have, say, -8dB as track gain value. The audio content in the track is still the same, and the player that recognizes replaygain tags will simply apply a -8dB to each sample on playback.
So anything that digital volume change can do(good or bad, subtle or extreme), replaygain can do. But nothing else is going on.
Now if you apply the gain difference to the file itself(like what mp3gain can do, or like you can do in foobar if you tell it to apply replaygain or other DSPs while converting a file), then the amplitude change is permanently applied to each audio sample in the file. But usual replaygain is just a scan and an extra tag in the metadata to tell how to adjust the level.