My point is that putting arbitrary restrictions on the mix is an exercise in the pointless. All of the "bad practices" you describe (which I agree with) are incredibly naive mistakes which will not be rectified by not touching the faders in the first place or chasing your tail by turning all of the mixer channels down as soon as the master clips. There is no need for this, a single gain stage at the master bus (fader changes on the master are still in the 32 bit floating point domain) is equal to (and far less hassle) turning down all of the other channels. If you don't believe me, try it: bounce down both versions and flip the polarity. The result? Digital silence. Keeping channels under unity is a good practice, but not if keeping the master under unity means having to rebalance your faders each time you change the dynamics of the mix.
I'm not quite sure where you see I'm coming from. I'm talking from an extreme POV. My point is that people who are not to sure about setting levels correctly before mixing will, more often than not, will try to make things as loud as possible by cranking the gain way to high on a channel, boosting the bass going mad with processing will result in a crushed mushy distorted dynamically dry mess, pre master fader. May not say it's clipping, but it'll sound bad. No point cranking down the master to compensate for a terrible mess.
Actually, clipping will not happen in practical cases until you step out of the hi bit-depth floating point domain and into the lower bit depth fixed point domain that the DAC expects, i.e. post master fader. And this is precisely my point: keep your mixer channels under unity, sure, but a gain stage is still a gain stage, whether it's at the master stage or across the mixing channels. Which is more efficient in terms of workflow: turn down mixer tracks each time a channel pushes the master over unity or simply pulling down the master? Pulling down the master, say, 3dB or pulling every other mixer track down 3dB?
Yes, nothing digital can go over 0db. Which is the main reason why it sounds bad! Values that sum to a value greater than 0db get chopped off. Period. This murders waveforms and you can see it as flat tops & flat bottoms. Where as in analogue clipping it still retain a rounded shape. So you may not be peaking, but you can still hear "clipping" from a) audio being crushed too much by processing (such as hard limiting) or b) audio getting chopped off It's not pleasant, and my point is being aware of your levels and setting them correctly. I rarely let my channels go above -2db when mixing I never let them touch 0db, and the I leave about -3 / 4db headroom before mastering the track. But that's how I like to do things.
Edit: I hope our conversation isn't off topic but at the risk of enticing a mod warning I'll summize my argument as thus:
1) there is no excuse for mixer channels going over unity. Although the channel will not actually clip, it's a bad practice as it's a misuse of headroom for nothing other than sloppiness. I believe we both agree on this point.
2) a gain stage at one point is equal to an identical gain stage further down the signal path, so long ad no dynamics-sensitive process is between the two said points.
3) Considering that faders are (almost exclusively) post insert, a mixer channel gain stage is identical to a master bus gain stage.
4) Now, here is the important bit: as mixer channels do not clip (actually clipping is irrelevant to this point but I'll keep it in this edit for clarity), comparing peak and RMS levels of mixer tracks is entirely relative. Thus, if no channels are over unity (for the sake of best practices) then any point of reference is valid.
5) closing point: if all perspectives are valid, there is no benefit to rebalancing mix levels to stop the master clipping. That is what the master fader is for. Thus, so long as you choose a sensible reference point, there is no need to compensate for master bus clipping at the mixer track level.
6) so, back to my personal method: I choose 0dB ad my ceiling when mixing. My kick will be at 0dB and not much else. Thus, mixer levels become relative to my kick, so any channel going over unity is a sign that headroom is being wasted I'd mix elements that should not be perceived as bring louder than my kick (I make House, hence the kick thing). I do not burden myself with balancing and rebalancing my mix to stop the master clipping, thus my workflow is not inhibited by such wasteful correctional processes.
Hopefully that's cleared a few things up for you about my position.