Probably the most poorly approached and understood topic in reloading. I am going to get wordy on this one. I will make two posts, one explaining the problem, the other listing some different solutions (though I do not think any are ideal).
I have had bottleneck rifle cases in caliber .233, .250 Savage and .308 Win. split on me. It depends on the brand of brass when it starts happening, but it mostly seems to come in the 3rd to 5th firing. I full length resize all calibers I reload.
Cartridge brass is 70% copper, 30% zinc, and is formally referred to as UNS C26000 brass. It is an extremely ductile material when its crystals are large and evenly formed. But, every time it is deformed while cold, its crystals likewise deform, and are broken into smaller, distorted crystals that don't allow smooth deformation. This process is called work hardening and occurs in all metals. As it proceeds, the metal becomes stronger, but less ductile (more brittle), and will eventually fracture under load, instead of deforming.
To undo this, we need to recrystallize the metal, to restore the large size and even shape of the grains. The recrystallization temperature of C26000 is usually given as 290 C, which means that if the metal is heated to this point, it will spontaneously recrystallize.
Take a look at the graph on page 29 of this document. It shows the effect on the hardness of brass of heating to different temperatures.
http://web.nchu.edu.tw/~jillc/me/Ch10-%20Kinetics%20-%20Heat%20Treatment.pdf
Bearhunter is annealing at 500 F in his oven, which is about 260 C. Looking at the graph, we see that he is barely starting to soften the brass at this temperature, and is probably reducing the hardness only about 2-3 points on the Rockwell 'H' scale.
Sawdust's molten lead bath is at some temperature over 327 C, and if he we assume he is a bit over that, he could be at 350C, which is the inflection point at the upper end of the recrystallization range. Figure he is dropping the hardness about 15 points, HRH.
Several people have mentioned heating the brass until it starts to glow red. All metals luminesce at around 700 C. Going back to our graph, these people are reducing the hardness at least 35 points HRH. Note that the people who made the graph discontinue it around 750 C.
For what it's worth, Matweb.com lists the annealing range of C26000 as 425-750 degrees C. This indicates that these are the temperature limits to what may produce viable results, usually in commercial practice. Lower than that, you are not doing much good, higher and you are making the brass way too soft to be practical, and risk oxidation attack which may destroy the metal.
Hornady is the only manufacturer to attack the issue. They offer an "annealing system", which is just a case holder and a bottle of 475F Tempilaq. Tempilaq is a liquid that looks like nail polish that changes temperature at a set point, in this case 475 F, which is 246 C. I applaud their effort to do something about the witch's brew of bad practices in the world of cartridge case annealing, but I really think they need to step up to a higher temperature. But, the fact that they are selling this and apparently having some good results with it adds to the data from Bearhunter and Sawdust that we really want to just be at the top end of the annealing range to get the best results, and the ideal temperature for home cartridge annealing is likely 275-350 C (525-660 F).