I must disagree with you.
While most of the necessary data can and will be aqquired by testing and simulations made "in-house", there are still things which cannot either be accurately simulated or tested by using the conventional methods.
While the effects of sub-zero temperatures on semiconductors are well known and can be simulated very accurately (basic physics), there are many delicate variables which still require real world testing.
The wide variety of data aqquired by the different testing methods will help the chip engineers to understand the behavior of the part even more better. Sometimes the data aqquired by the "unconvetional" testing might reveal some pitfalls or limitations which can be easily fixed in the next chip revision. These fixes can and usually will improve the 'experience' of an average consumer too
Still, I agree with you on one thing: "extreme overclocking" is nowdays mostly marketing.
In my case, the scale leans more towards to the side of development.
Of course it is advisable not to believe everything that you read from the internet.
I have tested three different A10-6800K specimens.
They are all from the same batch as the parts sent to the reviewers (1303SUS).
From what I have heard the retail parts which are newer production (1314 / 1318) generally seem to overclock even higher atleast on conventional cooling.
Two out of my three 6800K chips do 8GHz with both of the CUs enabled.
The other requires 1.925V and the other a bit under 2V.
The third chips does 7936MHz.
There is clearly evolution happening as with Trinity I had to go thru around 20 chips to find one that could breach the 7.9GHz mark.
Regards
Roger