Yea, for the vast, vast majority of workloads just forking separate node process ends up being better than mucking with threads.