ReactOS is entirely usable, if niche. GNUnet isn't usable. The home page describes it as "not ready for usage beyond developers", "still only suitable for early adopters with some reasonable pain tolerance", and "unlikely to provide good anonymity" -- it's unfit for purpose.

Hurd is an abject, embarrassing failure. Individual developers have produced more stable, complete operating systems as a joke (like PonyOS) than the entire GNU project has managed in 29 years.

Emacs is also a more complete operating system.

Does contributing to Linux not count? Some things are hard and ultimately fail. It's impossible to have any success without toleration of the possibility of failing.

Some of the GNU userspace projects surrounding Hurd have become important parts of the Linux userspace. Hurd itself, however, is wildly different from the Linux kernel, and has had little to no influence on Linux.

I think they should have backed out of it a little bit earlier. Being able to recognize your own failing might be as important as success.

Well, you have to fail anyway, so failing quicker might save you some time and get you to a good solution quicker as well. Understanding that should be a core part of any HNer at least. Probably wouldn't hurd the free software world either. (pun intentended)