Sorry I've not been more clear.
Gain staging does nothing for loudness. it is a technique in the analog world used to avoid clipping any of the many gain stages in the recording/mixing process. This process is no longer necessary in the digital world with floating point audio. The only place you can clip in Reason (a floating point system) is the final output, so it absolutely does not matter what levels you see between the input and the output. However, there ARE advantages to keeping all tracks around the same peak level.
I would say that many professionals use some form of system with regards to levels, many coming from the analog world where gain staging is important, and many who work full time find a "system" such as using a consistent reference level as I've previously mentioned simply makes things easier for many reasons.
Leaving the term "gain staging" behind since that's not what we're discussing here…
Using a consistent reference level is a starting point. To simplify, all tracks coming IN to the mixer hit the same peak level. It is ONLY at the output, the fader, where you make the final adjustments. As for your confusion over how adding a track changes things, consider this. For each doubling of IDENTICAL tracks (which can't happen in a real mix), you get only 6 dB overall increase. On average, each doubling will typically add 3 dB. Going further, some tracks don't always play, so lower this number even further. Additionally, many tracks will not need to be at their full level in order to create a balanced mix, further lowering the final output level of your mix.
As a specific example, if you have 24 tracks in your mix, you would have to add 24 MORE tracks to increase the mix level by 3 dB. Adding ONE more track will hardly make any difference. BUT if you were to increase ALL tracks you WOULD see an increase in mix level (but that's how far you would have to go to see a difference on the output in this scenario).
As for your suggestion of "working out the levels at the end", remember the reference level applies to signals coming IN to the mixer. Mix balances apply to levels going OUT of the mixer. So when you mix, THAT is when you are "working out the levels". Combine that with the small changes that even adding a signal full level track to a mix make, once you get your mix levels close you're going to be fine. For my work, I've found that using a reference level of peaks at -12 dBFS gives me a mix level that does not clip - that's the first goal of using a reference level.
The second reason to use a reference level is for when you add compression or saturation devices to a channel/bus/mix. IF you use a consistent level for ALL tracks, these devices will always "see" an input at their nominal level, which means at their sweet spot. All devices such as dynamics/saturation etc. have a limited range where they will function correctly, referred to generically as their nominal level. If your signal is too hot, there is no setting where you can reduce the effect. If the signal is too low, there is no setting where you can get ANY effect. Though it is unlikely you will increase or reduce a signal enough to exceed these ranges, by using the same level for all tracks you will make your job simpler since you will get a consistent response from these devices - thresholds will always be in the same range for the same result, for example. This simply makes your workflow simpler and quicker, and also allows you to save patches of your favorite settings and know they will work with little adjustments the next time you call them up!
To answer this question:
I still feel sure that there's more I don't understand. The mental image I have of gain-staging is:
1) on adding any track or instrument to the song (at the outset), make its signal match –12 dB, then
2) somewhat undo step 1 by reducing the level to where you really want it
Not always, sometimes you leave the fader where it is. But here's a question - where do you set the level of your FIRST track? Don't you have to start SOMEWHERE? And I'm suggesting CONSISTENCY, saying all tracks should be set the same, so I'm suggesting all tracks start at the same level for all of the reasons above (including some I've not had space to mention). My suggestion for a level of -12 dBFS is taken from experience, from other engineers, and even from the Reason User Guide! And not, every track is not the same level, because of where you put the faders to set the mix balances.
I said:
BTW, "headroom" just means the distance from the highest peak and the clipping point. Make sense?
You asked if it was Peak or VU level, but the answer was in my statement: "…from the highest PEAK and the clipping point", so PEAK level is what I was talking about.
Maybe start a new thread in the Music section where we can discuss a specific track, something I find can be valuable since you will hear different perspectives.
Keep your questions coming!