If I understand you correctly your comment boils down to whether or not you agree with the approach that Elixir and Erlang take to mutability.

In other langauges (even in Clojure, where I have a lot more experience) there are mutable global variables that you can use for these purposes. Unless I missed them then in Elixir/Erlang there are not.

Elixir/Erlangs approach to mutable state is the process. It doesn't matter what your specific application of shared mutable state happens to be.

We could discuss how any specific use case is more-or-less performant, but I would conjecture that someone who cared signficantly more about low-level performance over other things (e.g. systemic robustness) probably doesn't start with a BEAM language.