For example, with a stock install of Ubuntu 12.10, the user can not install the NVIDIA driver without crippling his desktop. Why? Because Canonical has been mailing it in for quite a few releases these days. Why should the concept of a "closed source" driver be blamed for what is clearly an error from the OS vendor? Lightdm was another situation. If it attempted to bring up the display manager before Xorg was loaded... kaboom... the user was screwed with a blank screen. How is that the fault of a closed driver? I had a fun time troubleshooting that one. In some of these cases it's impossible to believe that there was any testing at all.
I also take exception with the idea that NVIDIA is the weakest market player. First, they have the best Linux driver (fact). Second, they are actually in strong footing in the GPU market. They have far superior GPUs to those of Intel and AMD as a company isn't exactly enjoying financial bliss.
And, with all of the frustrations I experience with Linux (which fortunately aren't too often), it's hardly ever NVIDIA that I'm shaking my fist at.
Dare I say, the big problem with "Linux" (from a noob perspective) is that these distributions are all beta (or even alpha) releases. There is NO quality assurance being applied, and that's just blunt and to the point. Look at how many times there are versions released with show-stopping bugs and regressions that would have been picked up had a formal regression testing phase been followed. Even an LTS like 12.04 shipped with major cluster-f's. They have to slow down and not release a single change until it has been thoroughly tested. That's just how professional software works.