"There are a lot of things you can do with C++ (and other languages -- even C probably) that feel good at the time, only to learn later that they don't age well."

I think this reduces to the argument that you should do the simplest thing that works. As a programmer (especially as a C++ programmer), it's tempting to write clever code. It's fun. Why do something the boring straightforward way when you can invent an elaborate class hierarchy, or use that CRTP thing you keep reading about? Being a good C++ programmer requires a certain amount of discipline, but I think that's a trait that's helpful with more modern languages as well. Consider languages that allow monkey patching. It's a neat feature, but if used without discipline will lead to code that doesn't age well because it imposes more complexity on the codebase.