Hounddawg,
As I commented in another thread, I think we are at least partly on an apples-to-oranges comparison in a couple of ways. The shooter’s interest is in partial annealing (stress-relief) to make the brass more malleable and not in the grain growth that occurs with true annealing. In George Vander Voort’s paper, go down past the first seven rows of metallographs and look at the plot labeled Figure 10 on the right end. You will find it is not in disagreement with the other sources. It shows the Vickers hardness number for a 50% reduced area (RA) brass sample dropping dramatically when a 30-minute heat exposure starts to exceed 500°F (260°C). His text says he doesn’t see a change in the crystal structure in the metallographs until above the higher temperatures mentioned, but just because you don’t get a visible change doesn’t mean there is no measurable change. Indeed, this comports with what you found in the Matweb link. Note the Physical Properties table of grain growth (something you can see in a metallograph) starts at about 698°F at 1-hour exposure up at the top, but the Mechanical Properties table of tensile strength under it starts showing a change at just 392°F at 1-hour exposure (from the bottom up). The caveat here is these numbers only apply to the sample hardness they started the annealing with. The MatWeb sample was quite hard and past the 50% RA hardness, so the start temperature was lower. The loss of hardness and increase in malleability at these temperatures is what makes putting brass in a hot domestic oven, head and all, unsafe.
I can see Vander Voort used 50% reduction in area (RA) and 70% RA, but H. L. Walker used 70% cold-worked (CW) brass, which means 70% of the volume of the sample was cold-worked grains. The two correlate pretty well, but I am not clear exactly how well as the numbers get above 50%. Obviously, you cannot have 100% RA as that would be zero area leaving you with no sample to test. Below 50% RA % CW lags % RA, as your last paper illustrates. The author reports using samples that are 10, 20, 30, 40, and 50% RA. His last graph reports these samples as being 7.4, 15.7, 25.9, 33.8, and 47.2% CW. So, logically, you have to go beyond 70% RA to get to 70% CW, but I don’t know for sure nor how far. The cracking problem would make this difficult to do with rollers, but I’ve found one paper that goes to 90% RA by roller², albeit with some unusual treatment.
Here’s why what Walker found matters to recrystallization:
"In accordance with recrystallization law"…"—which states that increased strain on the metal decreases the annealing temperature—the time to reach complete recrystallization at 450°C {for} 21% CW {brass} was 4 hours whereas at 42% CW at {that} same annealing temperature it only took 15 seconds to reach complete recrystallization."¹
Those are not the CW’s or temperature we are concerned with, but you can see the principle at work. Merely doubling the CW produced a nearly 1000:1 reduction in the required exposure time to achieve recrystallization at that temperature. Obviously, if you wanted the 42% CW sample to take 4 hours to recrystallize like the 21% CW sample did, you would have to use a lower temperature. This is why comparisons hold either the temperature or the exposure time constant. By the time you get to 70% CW, the drop is greater³.
So, brass doesn’t have just one annealing temperature. How hot you have to get and for how long varies with how hard the brass is to start with. Ditto for stress-relief, based on the different tables. I suspect this may be one reason Litz didn’t find performance differences in cases annealed every load cycle. The brass wasn’t getting hard enough in one load cycle for the annealing temperature he had for it to make any difference. It needed to be hotter or longer.
The other reason Litz didn’t find a change is evident on the MatWeb page of properties. Note the Modulus of Elasticity (Young’s Modulus) has just one number for all of it. Look up MatWeb’s pages for different brass tempers and you will find it is the same number for all of them. It doesn’t change with hardness. This number describes the elastic stiffness of the material within its elastic (no permanent deformation) range. It is how hard you would have to pull on a unit length of it to stretch it a unit length. That’s not physically possible to do with metals without exceeding their elastic range (exceeding their yield point), but it is projected from smaller amounts of stretching because the number is used in engineering calculations. The fact this number is constant means that when you seat a bullet into a case with interference-fit, as long as you stretch the case mouth the same amount and it doesn’t exceed the yield of the brass, you will get the same amount of hoop strain (circumferential tension) and stress in the neck. That means you will get the same amount of force applied to the bullet to create friction in the grip on it no matter what hardness the brass has unless it is so extremely soft that insertion exceeds its yield, a number that gets lower as hardness drops and malleability increases. Short of that extreme, it leaves you with nothing different to affect start pressure or powder burn.
As to amateur annealing, if a case neck work-hardens to the point its cohorts are splitting, it will have its life extended by partial annealing (stress-relief). Splitting happens because the small percentage of stretch that occurs expanding it during sizing and seating and firing it becomes great enough to exceed its elongation-at-break limit, and a few percents of elongation can do it only if the brass is very hard. So If you can just get the neck and throat up past the 392°F point for any length of time, it will start to take the edge off the stress and reduce hardness. If you don’t get it hot enough long enough for complete stress-relief you will have to do it again more often, but incomplete stress-relief still lengthens the elongation before break number. This is why something like candle flame annealing works at all.
I don’t think exact partial annealing is critical because Young’s Modulus stays the same. The drop in hardness increases rapidly with temperature past its threshold. You can over-anneal a case to the point you weaken the brass and exceed the elastic range by seating a bullet, but other than that extreme, it’s going to be hard to tell from load performance. Just as with under-annealing, you wind up having to re-stress-relieve very soft cases more often to keep them from splitting because their elastic range gets too short. Elongation-before-break is the limiter at the low end and loss of elastic range (too low a yield point) is the limiter at the high end and are the brass life enemies produced by under and over annealing. Manufacturers aim to land between those extremes, and that seems to have a range to it that keeps the process from being too critical. Most people would like to achieve an ideal complete stress-relief without grain-growth because then they don’t have to do it so often to preserve their brass.
Note, too, that if you minimize case working and have a chamber with a fairly tight neck that doesn’t let the brass expand much, you may never have to anneal. If you can keep the neck expansion down to a percent or less, you may never exceed the elongation-before-break of the brass even when it is work-hardened.
¹
H. L. Walker. "Grain Sizes Produced by Recrystallization
and Coalescence in Cold-rolled Cartridge Brass,"
Eng. Exp. Station, 1945, vol. 43., p.7.
²
David S. Saunders, Report MRL-R-778, “The Low-temperature Annealing of 7.62 Cartridge Brass Cases: Stress Corrosion Susceptibility”, Defence Science and Technology Oganisation, Materials Research Laboratories, Melbourne, Australia, June 1980.
³
John Kline, The University of Illinois at Chicago, Department of Civil Engineering and Materials, paper CME 470, Recrystallization Behavior of 70/30 Brass, 2016, p 13.¶ 1. 
