The post dates the rewrite effort as happening in 2016-7. In the past couple years escape analysis has gotten a lot of effort (https://github.com/golang/go/issues/23109) and there's also now mid-stack inlining (https://github.com/golang/go/issues/19348) which obviously can be useful in itself but also help escape analysis and other optimizations work.
Does anyone know if there's any big picture work on GC throughput in progress? The idea of the ROC (fast freeing of memory never shared between threads on goroutine exit) was canned since it regressed on some workloads and I'm having trouble finding signs Austin's idea with hashing memory landed either. A moving GC or a read barrier seem mostly out because they can hurt other things (C interop, performance of code that doesn't allocate much). Though they seem to have addressed pauses quite well, and lots of code out there isn't GC-heavy, it still seems like a key area where it might be possible to give a "free" throughput win to existing code.