I'll try. But I would like to lay some groundwork first.

This topic does seem for many to be a quasi-religious. Meaning they seem to believe their one way is right in all places and times, and all who disagree are idiots or infidels. I'm sure they are nice people, but it is not worth my time to argue with them about this sort of thing, and I will ignore them. Nothing personal. You do not seem like such a person, thankfully.
Another problem, even in text arguments, is that many people don't read others' posts fully before firing off their response. So let's not have any of that here, OK?
I enjoy to argue (meaning "have a fair if passionate discussion") about ideas, but I do not like to argue with people. So please keep in mind that if I take a position for or against or around your idea, it is not personal. It may even be rhetorical; I may not believe it myself.

And if you don't mind, I'm going to take this in small bites. I'm also likely to challenge some of your assumptions, and try to present some of my own when they are relevant.
It's rather broader than simply that, but yes, this is really the main reason. Perhaps the only one. But it's a really significant one; how many such do you need? It has practical ramifications as well as engineering ones.

Rather than go into them now, let me start with this: For establishing a general rule, I am willing to trust in aggregate my named "authorities" over the relatively few exceptions that are usually offered as counter-examples. I am, in the same way, quite content to allow for such exceptions. When manufacturers make a specific recommendation for their cranks, I say you should follow it. I am willing to assume they, like others who may recommend differently, have done their due diligence, and have their own reasons for being an exception. General rules and recommendations are only general; they are not absolute. But they are to be followed unless there is specific contrary instruction.

As to the weight that I give to the consensus expressed by Barnett's, Sutherland's, Park Tools, Campagnolo, Shimano, and the JIS, versus Jobst Brandt, it seems so obvious I really feel little need to explain it, even though I shall. (Is there really anyone--who's not a manufacturer--who disputes a general rule of dry installation who doesn't either base their belief on, or justify it--incorrectly, if you ask me--with Brandt? Honest question. It always seems to come back to him or Sheldon Brown, who was basically quoting Brandt.)
The manufacturers and their organizations (Campagnolo, Shimano, and the JIS) all have literally decades of experience and millions of data points upon which to base their instructions. And it's not just "well, that's the way we've always done it." Most of the JIS bicycle specification covers durability tests and their protocols, not dimensions. And I say (with only mild humor) that it is pretty compelling when Campagnolo and Shimano agree on anything.
Barnett's, Sutherland's, and Park Tools, as educational organizations, are compilers and testers of each also have decades of compiling their own and others' experiences. They also all have close collaborations with the entire industry from mechanics to engineers.
In other words, any one of these sources has "big data," enough almost to make a general rule. Taken together, they have a mountain of it. It's a questionable mechanical practice to stack one's anecdotal experience against a mountain such as this. Mechanical things don't care what you believe.

Fools rush in, so let's deal with the elephant in the room. In the case of crankarms and bottom brackets, Jobst Brandt, for all his acknowledged genius and expertise, could not possibly have had a sample size that is remotely statistically significant compared to all these organizations' cumulative data. And as we all know (and many love him for,) he was not above using strong if occasionally faulty rhetoric to dismiss those who dared to challenge his conclusions. IOW, he was not always a good example of how to carry on a conversation. I think it's fair to say that his conclusions should be able to stand on their own without having to resort to his rhetorical stance, and mimicking that stance does nothing to settle his or anyone's conclusions. I don't dismiss them out of hand by any means, but I do feel compelled to take them as only a bit more than one very intelligent man's opinion, based on relatively anecdotal experience compared to the mountain of data of these other organizations.

Neither of those are conclusive, but for different reasons.

The crank-splitting argument is a straw man and I feel no need to counter it. Cranks crack from fatigue, not over-tightening. Any crank that does crack from over-tightening would have to have an internal flaw and/or be of very poor materials and engineering. The real problem is that we know nothing about Brandt's test protocols, if there were any, or his test parameters, or his sample size, or any of that science and engineering stuff. We have, as far as I can tell, only his relating of his experience as a conclusion. If we have otherwise, please let me know where I can read it.

As for Jan Heine's experiment, it appears to be a sample size of one, perhaps a few, and for a relatively rare sort of crank (namely with a forged taper.) Given that the only other crank tested was a Campagnolo C-Record (without dispute a "quality" crank) which has explicit instructions NOT to grease the taper, it is faulty to extrapolate towards any general rule about as vague a category as "quality cranks." (Jan does not specify whether he greased the taper for the C-Record crank, but if so, his experiment might support Campagnolo's dry taper recommendation....) Ultimately, Jan's advice is clearly meant ONLY as a specific recommendation for his Herse crankset, nothing more.

I'll take a break now, but will continue later.