I agree that the things you propose will probably converge much faster if you can get over the massive band gap. But this is very monolithic/centralized both in effort as well as control of data.
Mine is more of a distributed suggestion. We have a sea of monkeys available to us. Most of them rank Petarded on any intelligence metric you might to apply to them, aren’t particularly skilled, don’t possess a lot of means, and probably spend a lot of time on MCB. But there are a lot of them scattered about, so if we can distract them from spanking themselves long enough to follow some simple instructions, it might produce something useful. The monkeys need guidance.
If a test requires someone to have 10 cases of paint -- ANY paint, much less all of the same type, it is a dead test from the start. It is a total, absolute non-starter and nobody will ever reproduce that test. So you will get a max of one person to ever do the test, his word becomes canon, and the rest just copypasta. And everyone is completely hosed if that one person happens to be a manufacturer. That is the *last* group that you want to allow to control the data, yet by the construction of the test, they’d be the only ones willing to perform it.
This unfortunately has sort of been how the base of paintball knowledge has evolved.
Aaaaanyways, moving on.
The key point will be to catch the paint as gently as possible (genitalia optional). Again, that without the barrel, the velocity will be greatly, greatly reduced. Set up some cloth so it catches the paint just as it's about to start descending from the arc, like a landing ramp. It's basically like landing a huge jump in other sports. I think it's doable -- it’s just a matter of making it easy enough for a retarded monkey to perform.
Your #2 (huh huh) is going to happen incidentally by some of the monkeys just repeating the test in cold climates, so I don't think it's necessary to actually build it into the procedure. What's necessary is for monkeys to actually to just do the test and *record* their conditions. Otherwise you have to add temperature control as a barrier to entry to the testing, which again is not helpful.
The point is to just have people collect the data, correctly, with whatever they have, and then aggregate the data. Over enough data, trends will emerge.
It's more important for people to just record their conditions and setup, most of which are taken care of with just... a video. Record the test. For example, even if they don't record ambient conditions, if they state date/time and location, we can at least hit the ol' weather almanac and get an approximation of conditions. We can also see if they constructed the "catch" correctly.
If they screwed anything up, unceremoniously eject their data and move on to the next monkey.
This will be the path out of the fog. It’s either this, or not at all.