about time

However, AFAIK golang's schedule is still quite *not* a thing to be praised and admired the moment you actually need high-throughput AND syscalls. I am pretty sure we will need to force GOMAXPROCS around at work in the k8s clusters (where we typically need at least a 2:1 go-max-procs to container-cpu-cores ratio for a decent tradeoff of latency x throughput). YMMV and all that.

The GC and std lib improvements being done lately were much appreciated, though. They help a *lot*.