I have always been sceptical of the thought that certain batches perform better. Knowing how the chips are made, the varience between chips on the same wafer can be huge, even two that a litteraly right next to each other on a wafer can have huge differences. Its just luck of the draw, and with how many bins these chips have, it makes it even more difficult to believe that certain batches are better than others. Remember, the same batch can become a 3960x, a 3930k and a 3820.
Also, keep in mind that what intel believes is a better chip isnt in line of what we think a better chip is. Many people assumebthat ifnit can overclock very high, it is a supperior chip. This is just not true. Yes, it does mean that the high oc chips handle voltage better but that doesnt not equate to performace. Scaling is a much more important figure when it comes to oc than just raw values. A good sample of this gflops, errors and time to complete a task. In my experience, i have had chips that hit 5.4ghz but lose out ob scaleablity to chips that can only hit 4.8. Also, there are chips that hit 4.8@ 1.35v that cant oc any higher, but a chip that hits 5.4 needs 1.42v for 4.8.
To me, it is way more complex than just saying "this chip oc better, so its a better chip"
You make some awesome points, but I should point out that the 3960X and the 3930K are based on an eight core die, while the 3820 is based on a separate four core die. Your point regarding the multifactorial bnning process holds however given that C2 eight core dies from the same wafer could be made into anything from a 2GHz six core xeon to a 3960X or even a 3.1GHz eight core E5-2687W.