One of the many huge problems with v3, however, is that they opted to combine a complete rewrite of the server with a complete rewrite of the client AND a complete UI overhaul.

I don't know what the wire protocol for v2 looked like (probably crap; everything else was), but I struggle to imagine a scenario where they couldn't have rewritten the server code to allow scalability without needing to touch the client, or at least with minimal changes. (And let's not forget that v3 scaled like crap too, at least at first.)

Mind you, Wizard's wouldn't be the first (or the thousandth) team to inherit a crappy code base, throw up their hands, announce that they simply couldn't possibly salvage it, and ask for permission to rewrite it from scratch. It's just most successful places learn to say "absolutely not" well before they're operating at the scale Wizards is. (Or they fail miserably and go under.) I think what makes MTGO so fascinating is that's it's one of the worst software development disasters I've ever seen, and yet it's still a success...of sorts.