Flint,
Thanks for your kind words. Verifying stuff is both enjoyable and a priviledge to have the time for.
The product was only designed to smooth the surface with a hard layer, not to change anything about it's stress - surface or internal. I recall from my reading that the reason glass breaks so readily is due to this inherent internal stress, so I'm with you there. However, preventing cracks on the surface is all that is needed to make a windshield impervious to abrasion and chips and that's what they attempted to do. Too bad it didn't work.
As to the spherical thing. A random shaped rock has protrusions, flat spots, and even voids comprising it's surface. At a given speed, it's force in PSI could be 10 (such as if a pointy part hits the windshield), or a 1 (such as if a flat spot hits the windshield), but the average PSI of having that rock hit a windshield 100 times would be 5. Might take 1000 hits to statistically average a 5, but eventually it will. A perfectly spherical rock of the same size will provide a PSI on the first hit of 5. On the second hit - a 5. And so on. It's perfectly consistent and uniform in shape. So, testing with a rock would provide random levels of damage and with a BB it provides uniform levels of damage.
Translating that into my real world test, if I propelled a rock at the windshield and caused a chip on the uncoated area I would not know if it was due to the speed or due to an unlucky hit of a pointy part. Doing it again at the coated area and not causing damage, I would again not know if the lack of damage was due to the coating's effectiveness or a lucky hit on a flat spot of the rock.
So, the BB was the best in terms of a repeatable test. And it used a real windshield and a pea gravel sized object actually striking the windshield at high speed.
Speedwise, the only "certain" speed we knew was that the cheap spring type BB gun was rated at 350fps. Really moving. The air gun felt much slower at one pump but it was an unknown speed. So, we focused on using the BB gun since it was A) a known speed, and B) more consistent in speed.
In actual testing, we backed up so far that we were certain there would be no damage as the BB would certainly slow from a way too fast 350fps. Though we really didn't care what the speed was as long as we found a distance to fire at that had enough energy to damage the uncoated area, but that would not damage the coated area. That's a key point. Statistically, our theory was that there was protection provided by the coating.
At 100 feet we were unable to damage the windshield, held at a car like angle in a wood/foam bracket that did not allow for any movement. After 20 shots - not so much as a mark. At 87.5 feet (halfway to our 75 foot marker) again no damage on either coated or uncoated. At 75 feet, the first shot chipped the uncoated area and we were pleased we'd carefully found the sweet spot, anticipating that the coated areas would be impervious. Alas, first shot into the coated area chipped it also. So, we stayed at this range peppering the windshield and tallying hits/damage/zones (we divided it into 3 squares of uncoated labelled 1,2,3 and everything else was coated). No statistical difference.
We even noted that hits on the curved part of the glass tended to be more resistant and did not damage the coated area. Alas again - we could also not damage the uncoated areas on the curves. So again - no difference. So, it was a reasonably controlled, repeatable and accurate test in my opinion. I staked an investment upon its outcome and have absolutely zero doubts about its accuracy for my purposes. It was at this point that we became curious just how much damage the pumped up air rifle could dish out and limbered up with it at 15 pumps. We were all stunned by its power and only did that once.
The sandblast test was similarly accurate, controlled and repeatable though it was over in like 3 seconds. Sandblasters use fairly consistent and clean graded and sifted sand, and they're all moving at the same speed. A single pass over the glass (I covered up 3/4 of it in case we screwed up and were too close so we could try 3 more times) gave it a lightly pitted finish exactly like a windshield around here looks. But no lines appeared separating the coated/uncoated areas. Zippo difference, so we stopped the testing immediately.
As I said, this was disappointing to me. But I'm glad I did it as I would have been not only extremely embarassed when customers started returning in droves, but would have lost some accounts as well.
Regards,
DougM
Next couple weeks from now is going to be my 4th f&*%in times I replaced the LX windshield with rubber molding b'coz of the stone chip, cracked... 
