I think the problem then is...if you have some promising directions from preliminary work, to test an idea in clinical trials, it is not cheap. You have to hire a large army of clinicians, nurses, research people, etc etc to manage all the patients/subjects x 10,000 or 50,000. So there has to be at least some politicking to steer where all the research funds go to. And people who are better at "finessing" scientific review boards tend to bubble to the top
If the fundamental problem is misallocation of research dollars, then that's what should be tackled first. I'd vote to do something like a git bisect and isolate the failures first. The scale of the failures suggests that the problems go back to the beginning. So I'd start by halving the funds that go to the big players and distributing the remainder amongst smaller groups and individuals trying a wide array of different approaches.
That said, I'd predict that as theory and simulation improve, especially in areas like big data, machine learning and AI, strong correlations will be found soon (probably within 5 years, certainly 10). They'll probably discover something that the article hints at, that the problem is actually inside cells and that the amyloid protein is a symptom (not a cause). It's probably something like a subtance in food confusing the immune system or a multi-cause effect related to living a lifetime under acute stress far above and beyond what we'd encounter while living in a hunter-gatherer society, or simply that we get no exercise compared to what our bodies evolved for so maybe the cellular repair mechanisms generated by our muscles aren't there to repair the nervous system (no blame here - it's hard for older folks to get enough exercise when they're hurting).
All wild speculation on my part as a computer geek who doesn't know the first thing about medicine, I'm the first to admit!
The thing is, the beta amyloid theory was the only explanation for Alzheimer's that would allow treatments to be developed right now.
Tons of people want treatments asap.
While it makes no sense to develop treatments based on this theory if they definitely won't work, when it looks like there's an 80% chance of the theory being wrong, 100% of drug companies' Alzheimer's drug development money is still going to go to drugs based on this theory because we have nothing else except vague hypotheses with no drug targets.
So actually part of the problem is that so many people are incentivized to keep sticking to this theory and someone needs to step in and stop it.
It just have a longer life than normal ponzi schemes, but ends up in the same way in about 20-40 years.