Over the years I have seen several people make comments about OpenGL compatibility issues on Windows. Just little things here and there. I've never come across a specific problem. Is it FUD or is there some truth to it?

I know that if you code on an Nvidia card, that the Nvidia driver is very accepting of bad OpenGL. If you write code using an Nivida card, then try to run it on AMD GPUs, your program might not work. This leads me to believe that one should write OpenGL code using AMD GPUs since those are more strict about requiring proper driver calls.

Other than Nvidia's crap with their automatic OpenGL driver fixes for bad code, are there real issues with OpenGL on Windows?

Don't know which issues you are referring to specifically. But the #1 problem most people will encounter with OpenGL/ES which really doesn't depend on the OS is the inconsistent behavior( which can be interpreted as incompatibility). Most people including myself will agree that the Nvidia drivers are less strict in adhering to the specification, but assuming that AMD is correct 100% of the time is also wrong. I've personally found and reported a bug or 2 is AMD OpenGL drivers ( 1 specifically in regards to subupload of compresss texture, which strange enough their core function exhibited incorrect behavior, but their implementation of a Nvidia extension for the same functionality did not) . The inconsistencies are even more greater on the OpenGL ES side of the house as I have seen different behavior from PowerVR, Mali and Adreno GPUs. Inconsistencies are somewhat inescapable due to how the specification are written. If things are left open to interpretation then its assume that liberties will be taken that will benefit the IHV hardware. If the underlying software is written with certain implicit assumptions instead of checking for most eventuality then chances of running into an edge case in terms on OpenGL increases.