Mixerman, I find that your statement " the sound gets smaller the more sounds you put into it" rings a chord in my brain.
The really interesting thing is, lemme see if I can put this into words :)

When assembling stereo final tracks, like, in order to crossfade from one song into another, or mixing two stereo signals together ( in both of these instances I speak of the already final finished mixed tracks ) I think that the summing in the DAW is undeniably better than trying to assembly edit on a console .
First, there is no aditional conversion hit, or generation loss.
Second, there are no unnecessary passes thru amplifiers.

If you scope the output while you are doing this, the waveform during stages where there is only one ( i.e. just one waveform playing at the moment ), even though going thru the mix bus of the DAW, looks EXACTLY the same as the waveform on DAT. Of course during the mixing it does not.
This only happens if you are careful to find the "zero settings" for each DAW, many times the manufacturers are iffy about this...SHAME on them!

Anyway if a DAW can mix 2 stereo tracks together just fine, why cant it do more?

I agree with you that it gets worse as more tracks are added. I would prefer that it be truly " garbage in garbage out " and just turn things down so they dont clip...I dont think thats so much of a problem at 32 bit anyways! Do manufacturers think we are too stupid to turn it down when we see clipping?