Swap out the global allocator for something faster (especially for multithreaded scenarios) like: - [`mimalloc`](https://github.com/microsoft/mimalloc) - [`rpmalloc`](https://github.com/mjansson/rpmalloc) (potentially slower than newer mimalloc in benchmarks) Might be good to build one into Faabric with strong symbols for `malloc`/`free`/`realloc` to replace the glibc-provided allocator.