Would be interesting to read some more details about how this got through, and what the pitfalls were, along with solutions to them. As I understand it, the main tension is that things that look more like industrial consortia, developing common technology primarily for the benefit of an industry, don't qualify as charities, though they may qualify as other kinds of nonprofits. For example, the Khronos Group, steward of OpenGL, is an industry consortium and not a charity. While organizations primarily aiming to educate and serve the general public, rather than primarily to improve the technology or standards in an industry, can be charities. LLVM as a project seems—to a non-lawyer—like it could go either way, since it has elements of both types of goals.