Obviously these articles have all been conceptual so it has to come with a big grain of salt, but if we presume perfect execution it sounds like a great idea. Of course, perfect execution is improbable and so problems will creep in.
Well, yes, perfection is impossible. However, this would be true of any RPG system, regardless of complexity, or "dialed complexity".
The first problem I see is that I can't imagine a way for this to be developed where you are producing the best systems possible. You'd have to make sacrifices in the design in order to ensure something could be easily turned on or off. This means that when someone wants to flip the skill-switch 'on,' they aren't getting the best skill system D&D could design, they're getting the best skill system D&D could design that you could also choose not to use. They are pitching a game that appeals to the widest possible audience (which is smart) instead of the best game for the audience they have (which is a bummer).
My other concern is that creating one system of D&D that plays so differently will diminish the ability of the community to get together and dish. Talking about the game is a *huge* part of the experience, and if 5e looks like four or ten or whatever different games based on what you flipped on or off, it might make it harder to talk about. That'd be a real shame and I think I might add "the ability to quickly understand the play experience of other gamers" to my list of D&D sacred cows.
I think this type of "variable" or "dialed" complexity is perfectly possible, and I'm excited to see where this goes. I think the answer to your concerns is that while the complexity options would potentially be on a continuum, there would be a small handful of "sweet spots" in the design. You a fan of the original D&D or the BECMI D&D? That's "sweet spot" #1, with very little in the way of customization for characters and game rules. This could even be marketed as the "Red Box" or Basic D&D.
You a 4th Ed fan? Like plenty of customization options, but like things to still play relatively simple? Sweet Spot #2. Sweet Spot #3 might be akin to 2nd Edition with Player's Option. You can customize your game further inbetween and around those "sweet spots", but the designers would provide solid packages to cater to specific play styles.
And when talking about the game on ENWorld or other forums, we all may be playing at different "sweet spots", but the underlying rules are all the same. And your monster build simply and my monster built with tons o' options could still play nice together and we could share the ideas and statblocks. Not unlike how it's easy peasy to play a 4th Ed "classic" character and a 4th Ed "essentials" character in the same game with NO problems. Heck, back in the day we used to mix BECMI D&D and AD&D all the time . . . it wasn't perfect and we did it because we didn't "know better" . . . but you know what, it worked.
Of course, this type of system could be designed badly and be a huge failure, or it could be designed quite well and be the version of D&D that brings us all together to sing happy songs while we slay orcs . . . but that's true of any design approach.