Usually, when someone asks to get away from something that's in widespread use, well-tested, verified on many platforms, it's an outward expression of an underlying problem known as "code smell" and the uncontrolled accumulation of "technical debt" or "code debt". The GNU archive had built up a fairly large amount of code debt over the years, and when a codebase is not properly maintained it can reach a breaking point (legacy code, and even morbid legacy code).
Normally, one would carry out a process of re-engineering and refactoring at intervals to keep this under control. So, the real question that's being posed here is whether a refactored version of coreutils has been developed. This, of course, includes the possibility of an outright replacement (as a special case) - much like Wayland is being pitched to be for X ... many of its developers coming straight out of the X camp.
My suggestion is to actually go in and refactor coreutils. Someone's gotta do it. And whoever raises the issue of replacing coreutils - your idea your project.
To this end, take advantage of whatever automation you can find: refactoring engines, like cscout, or anything that applies more advanced analysis/synthesis methods (e.g. formal concept lattices). But deep analysis is still a relatively new and open area of active research - and crosses over into Artificial Intelligence. (A robot software engineer.)
Most of the utilities should already have test suites in place, so validation can be done with progressive step-wise change + automated regression testing steps; which can go pretty fast (e.g. 10 or more revision updates/day). A complication to this process occurs if there are hardware or low-level software dependencies anywhere in the software suite; since that entails validation on multiple platforms. I don't know much of that there is in coreutils; there should be some kind of separation in it from the hardware or low-level software layers (e.g. the number of places where coreutils knows what type of filesystem it is on should be minimal or, better, zero.) Emulators and virtual machines, put to use for purposes of doing multi-platform testing, have limitations. For instance, Mac OS X is specifically designed in a way to obstruct the ability to emulate or VM it.