Certainly, copy variation is a factor, since most reviewers only test one copy. A few years ago, Roger Cicala (LensRentals) would test 10+ copies of a lens model on a very expensive machine that took the camera out of the equation, showing standard deviation, etc. You really had a good idea of what you were buying and the relative odds that you would receive a good lens.
Having said that, it seems as though manufacturers have really tightened their tolerances, in recent years. It used to be common for me to return a lens that did not pass my tests, but it is now rare that I encounter such problems (as long as I am not too picky about de-centering).
In December we took a trip to Europe. I didn't want to carry a lot of gear, as I usually did, so I purchased the newly released Sigma 20-200mm. My expectations, to be truthful, were not very high. When I got home, and in reviewing my RAW images at 100% on a 27" 5K monitor, I was almost stunned with the sharpness of this lens in the central area. Tiny details, when blown-up, were amazingly sharp - similar to my GM lenses.
I then looked at the tests on LensTip.com which showed an unusually high level of sharpness for this lens, so I don't think mine is a fluke. Also, the contrast was excellent (much better than the Tamron 28-200mm lens that I had).
I highly recommend this lens not only for the IQ, but for the ability to go down to 20mm, which makes it highly versatile when you do not want to spend a lot of time switching lenses.