I've been following the discussion so far, Moose. All i get from your wish list is something that was explained over and over why Second Life can not afford that, but you keep insisting. Your two points don't make sense at all, if it's not a professional platform, then why you and all the other creators who pump the models so much act like if you were? LL doesn't provide a efficient tool? Professionals buy the tools they need, and MooTools Polygon Cruncher is excellent for that.
Maybe Optimo is a bit stubborn and harsh in making his point across, but he's right. I talked to him last night and it's a matter of math used to convert the mesh text file (.dae). The text file has no limitation for what concerns values and numbers, but the binary conversion is limited to a specific range. And he's right saying that Lindens have stated more than once that they wouldn't mess up with the mesh asset architecture.
So we all get your point of wishing something from the Lab, Optimo himself said he would like that too if he didn't know what impedes it. What i would wish more than a raise in vertex limits, though, is a more relevant weight for the lower lods. Low and lowest vertices count affect the final LI too much which, to me, doesn't make much sense.