That's interesting! Would it work to have a process on macOS/Windows monitoring memory usage and communicating with another process inside the VM to drop the page cache when free host memory is low?