Rich replaced the default musl malloc some time ago for exactly those reasons. Maybe they still used the old musl libc?

The new one was drafted here: https://github.com/richfelker/mallocng-draft

The new allocator does nothing to improve the performances in a threaded / contended application: https://www.openwall.com/lists/musl/2025/09/04/3

The response to the link here is really telling.

Blames it all on app code like Wayland

From the article:

> “the new ng allocator in MUSL doesn’t make a dime of a difference”

Yes, sorry, missed that at the very end.