The ideal solution would be to distribute high dynamic range audio with metadata to configure optional playback-time dynamic range compression for noisy listening environments or weak playback equipment.
Or make a sound format (like video containers) that could have two separate mixes of a track.