I'm thinking about it myself.. I havent bought a nvidia gpu in well over a decade but I think thats about to change..
Mostly people are doing something like flashing-in a custom bios/overclocking/overvolting/running too hot, etc., ad infinitum--and cards go "pop!" at some point. But to be honest about it, I've actually never lost a single 3d GPU from any manufacturer that simply expired (ATi, AMD, nVidia, 3dfx, STB, and others.) Never lost a single card. Came close, however, with a 3dfx V5 5.5k circa ~2000-2001 or thereabouts...;) I had changed heat sinks and was overclocking the crap out of the card--when suddenly it just went blank--wouldn't put out a signal of any kind (VGA or otherwise.) Put the original heat sink back on, and had rebooted about ~6 times consecutively and was preparing to give up when suddenly--ta-da--I got signal! System booted, and never had a moment's trouble afterwards. That was the closest call I've had since I bought my 3dfx V1 circa 1997-8?. But I probably don't use a GPU more than 18 months-2 years max, if that long. So, I suppose just plain longevity could be a culprit.
Don't see how just a set of software drivers could burn out a card--even if the drivers grossly overclocked the card somehow--it would just lock up as opposed to dying, and would unlock as soon as clock/temps got sorted.
I do know, however, that as GPU manufacturing processes get smaller the chance for electromigration damage from overvolting and overclocking rapidly increases--but the same thing happens with cpus, too. When you get to 14nm best observe stock voltages and cooling much more stringently, imo, if you want the processor to last.
After I installed the driver, this was a couple months ago, so it was one of the oldest crimson drivers maybe one of the first ones put out.. I noticed my r9-290 hit its temperature threshold and throttled itself, it got hot enough to do that, so an older card could definitely get hurt or toasted.