Yup, it's worth discussing. My nurses-talking-about-heart-surgery comment wasn't a dig (seriously, I know this stuff is fun to discuss). But... holy shit is this a tough problem. Basically the first sign that you (the proverbial you, not you specifically, of course) wouldn't even be able to make an impression (let alone a serious dent) is if you catch yourself thinking of the solution in terms of existing building blocks. I know that sounds harsh, but, from a programming point of view, if something like a completely from-scratch distributed storage layer is an amount of complexity/work that you'd prefer to avoid (by trying to take advantage of something that looks suitable and has already been built), then you've already demonstrated that (in all likelihood) you don't have the software engineering chops to handle the rest of the complexity involved. I mean, maybe I set the bar unnecessarily high in my previous post, by imagining it as an in-place upgrade with nothing lost along the way (feature-wise, UI-wise, every last bit of the database comprehensively preserved, etc.), but if you relax those constraints and simply say: "Can you make a decentralized forum similar to Bitcointalk that might one day be able to serve as its replacement?", then I'd estimate that to be a much easier lift.

(But, I've been programming since I was an ankle-biter, and all my years doing it have left me with a pretty depressing view: I think programming, as a profession, is regressing, and most of the legends have already left the building, so to speak. So, even in the watered-down version of this problem, I'd still expect most every attempt to end in failure. If not outright failure, then the slow, eventual failure that comes after you basically force a win by hiding the fact that you're dealing with more complexity than you know how to tame.)