You could take a look a swift-nio (https://github.com/apple/swift-nio) which is a pretty high-throughput system. swift-nio does this using some reference-counted GC where it simplifies the code and doesn't affect performance. Otherwise, value-types are used which incur no GC overhead (unless they are copy-on-write, and backed by something that requires reference counting).

Well that's why I said high level application use cases and not systems. Ideally I'd love to see a compile mode for modules where certain conveniences were enabled like: auto-clone, unified reference/owned types, disabled lifetimes and instead borrows become auto-wrapped-arcs, etc. That way you could care about that stuff in modules where it matters, and ignore it in areas where you prefer convenience over zero-cost abstractions and raw performance.

Sure you can. C++ has reference-counted garbage collection with its smart pointers and RAII, just like Swift. People certainly write high-throughput systems in C++.

Swift is slower than C++, yes, but not because of its memory management scheme.