[bodyfour]

If you have a single-threaded runtime then reference counts are cheap and easy. Once you need to support multithreading you need to use CPU atomic instructions to maintain the reference count safely. As the number of CPU cores increases the relative cost of this operation goes up.

COW strings were a very common optimization when 1-2 CPUs were standard. That's why it's there in GNU's libstdc++ for example. On a modern server with 24 cores it's pretty dubious. It'd only make sense if you were copying (and then not mutating) a lot of your strings.