Edit: There is a flaw in the method I am about to describe (and perhaps others). If you have sends on any of the tracks then read the following post (well read it anyway if you are going read and use my dumfuk advice because I don't wanna mess anyone up).
http://forum.cakewalk.com/FindPost/3286229 OP...
==============================================================
The way I do it is this...
I always insert a "Premaster" bus before the actual Master. All tracks get sent to the Premaster (by way of all the sub busses I create for everything). So it goes like this... Tracks > Sub Group Busses > Premaster > Master > Physical Outputs.
I keep my Master bus completely free of any effects and I keep it at 0db Unity gain as I work. I use the Premaster for any master type effects and to adjust levels if need be but generally the premaster is there just to test how my mix will respond to compression/EQ or other mastering type stuff. I am trying more and more to leave it completely clear and rely on the mix then just turn on/fiddle with those effects to get a rough idea of how glue compression and mastering EQ may go down once I get to the mastering stage.
Since my Master bus is totally free and clear of all effects and all my tracks go to the Premaster it creates a situation where I can route my reference track(s) main output to the Master and absolutely nothing will be altering the sound.
So then all I have to do is keep the reference track muted. Then when I want to compare my current mix to the reference track all I do is SOLO the reference track (which mutes everything else in the project) then unmute it. So it's maybe a one second silence at most as I click the solo/mute button. Not an instantaneous. one click a/b but it really isn't that far off.
Since I keep my general output level from the premaster (going into the Master) at around -6db I just have to make sure the reference track is outputting to the Master bus at -6db and it gives a proper relative volume comparison.
I actually keep previous mixdowns of the project and reference tracks in the same track within their own lanes. That way I can just solo the lane of the ref track/previous mixdown I want to compare with and use the method described above. This can cause a problem if my exports were -6db exports (for premaster) because then my ref tracks are of course going to be louder than the -6db exports (so with the 6db I removed to get the ref tracks matching my premaster bus level those exports end up outputting at -12db) but that can be worked around by putting the -6db exports in their own track and not lowering the level at all or leaving them in the main ref track and using clip automation to compensate for the volume difference).
It's a pretty decent system but I do need to spend some time with Exclusive Solo which I think might make things even smoother... and not just for this application.