There is also the problem of camera-body quality control. What I find curious is that my Sigma 35/F1.4 doesn't require any autofocus fine tuning at all on my D800E, yet it requires the maximum adjustment of +20 on my D7100.
On the other hand, my Nikkor 24-120/F4 zoom doesn't require any autofocus fine tuning at all on my D7100, yet requires a modest amount of adjustment on the D800E.
I'm thankful that the maximum adjustment appears to be sufficient to allow me to get sharp results using autofocus with the D7100 because it's too late to return the lens. You can see from my very elaborate test procedures in attached images that the +20 adjustment appears to have done the trick. The images are at F1.4.
Unfortunately, this terribly sophisticated test target is not for sale.
I don't own the Sigma lens but have been interested in buying it. Two years ago I bought the Zeiss 35mm f2 in Canon fit. It is an exceptional lens in every way except one. It's manual focus, and as I like using it wide open quite often, I do have trouble focussing the damn thing. In live view mode on a tripod it is great - but hand-held I do struggle. Hence considering the Sigma.
Ray, I did read an interesting article a while back on the subject of manufacturing tolerances. Your problems are just to be expected in regard to different lenses on different cameras - It's just a matter of tolerances, and how they interact with each other - some you win, some you loose.
And to some of the other comments here regarding problems with the Sigma lens - you just have to realise that it is built to a budget and will not compare to say, a Leica lens. Not that the lens is low-quality, just that there will be a greater range of tolerance accepted by quality control. And curvature of field affecting a lens at different focussing distances, de-centred lens elements, and focus-shift when using very wide aperture lenses are all problems for most lenses - not just the Sigma.