Quite a few people have reported that their problem is either fixed or reduced when replacing hardware. In my case, I only had to un-cripple the cooling infrastructure (haven't had a freeze since I did that). The cooling infrastructure, which is, by the way, the same across many versions of the HW.
If this were a pure software problem, then you'd expect everyone to have the problem, unless the variability came from something the users are *doing* or *installing*. I see no signs of a pattern there.
HW/SW *interaction* problems are quite common in computers, since hardware has a TON of variability that we usually would prefer to ignore. SW changes can easily unmask something suboptimal in a HW design.
ALL THAT SAID, your problem may be different from that that others are having. But in a substantial number of cases, changing HW and/or SW can reduce or fix the problem. That's the classic sign of a HW-SW interaction.
My guess is that Yosemite is switching too aggressively and that this is causing a local thermal ramp that's too steep in some combination of {GPU, CPU, Platform Controller Hub}.