If every single 24-bit digital signal would PEAK at -6dB, the world would be a better sounding place.
One of the big reasons - Use your imagination here -
Think of an analog sine wave - Now digitize it and run it up to 0dBfs. Simple right?
NOW, play it back through a D-A converter... What happens to the \"curve\" of the signal now? That's right, it's clipped. The analog signal, freed from being simply ones and zeroes, follows the analog rules. The peaks will go below AND ABOVE where the digital signal \"plotted\" the path. There are no straight lines connecting the dots anymore. The now analog (and naturally curved) lines can exceed the plotted points by quite a bit.
Superior D-A converters handle this with much more style than lesser converters (like a standard CD player or less-than-stellar sound card), but the distortion and the \"guessing\" is still there.
This is just the tip of the iceberg on the \"idiot's version\" of the physics behind this. It's about all I'm qualified to explain.
There are some great white papers and studies on this. If I can find some links, I'll post.
If it makes any difference, I commonly get e-mails (from people who used to try to get as close to 0dBfs at every stage) saying what a HUGE difference this makes in the sound quality of their recordings. From my own experience (I used to worry too much about \"0\" also), I can testify to this also. The open quality, clarity, punch - All greatly improved in the final product.
Of course, after mastering, few would want a disc that had a top level of -6dBfs... But when you think about it this way, if everything is lower and therefore less distorted during this final phase, then end volume will likely be able to be pushed farther than if it was distorting the M.E.'s converters in the first place...
Better-sounding at high volume... What a concept...