One of the fastest web servers is nginx that uses child processes (typically one child per core) and asynchronous IO within the child without any threads. The same model with a parent process that monitor threadless child workers is often used in embedded system. One can and people do exactly the same in OCaml.

Surely it requires more code to setup things especially if the child processes need to communicate over shared memory to max the performance, but the big plus is that the resulting system is much more robust. In case of troubles one can just let a child die or even use kill -9 and start a new child. This is not possible with most threading implementations. For example, try to recover from out-of-memory in multi threaded C++/Java/Go etc. application. It is very hard. With threadless children it is almost trivial.

Plus we have learned the hard way that concurrency and processes are much saner from safety point of view than threads.

Posted a week ago: http://kcsrk.info/slides/mcocaml_gallium.pdf

Discussion: https://news.ycombinator.com/item?id=17416797

In short, expect it merged in mid-to-late 2019, maybe.

While this is an issue, there are cooperative threading libraries (mainly Lwt and Async, though the community prefers Lwt) which can mitigate most of the reasons you'd need multithreading. Hell, because it's only single threaded, you can mostly ignore locking.

Isn't this project aiming to solve that?

http://ocamllabs.io/doc/multicore.html