I got stuck on sorting too, was working on SingeliSort (https://github.com/mlochbaum/SingeliSort) for a while. The basic performance is there but I need to get serious about testing before using it. But the radix sort and counting sort should be very solid. The approach is about the same as the C code currently used in CBQN, linked below. The main complication is to reduce constant overhead for shorter arrays with a small count type and better prefix sums, interleaved SWAR for SingeliSort since it targets generic architecture and shared SIMD utilities in CBQN.
Email in my Github profile, feel free to contact me any time if you'd like to talk about algorithms!