In http://docs.python.org/dev/extending/newtypes.html, you can read:

“To enable object creation, we have to provide a tp_new implementation. In this case, we can just use the default implementation provided by the API function PyType_GenericNew(). We’d like to just assign this to the tp_new slot, but we can’t, for portability sake, On some platforms or compilers, we can’t statically initialize a structure member with a function defined in another C module, so, instead, we’ll assign the tp_new slot in the module initialization function just before calling PyType_Ready()”

But the thing is, we ourselves (CPython) do exactly what is discouraged here, both in built-in types and dynamically loaded extensions.
So is this piece of advice still necessary?

The advice is still necessary, AFAIK. The issue is Windows, in particular producing function pointers across DLL boundaries. In Python core, this is not an issue, since the references will all be inside pythonXY.dll.