Finally some good news :)
In latest previews (>= 2014.4.0-0280) the SkiaRenderer node uses ANGLE for rendering, WGL_NV_DX_interop2 is therefor no longer needed. It uses the same Direct3D11 device as provided by Stride and (contrary to what I said further up and in the chat) also renders to the given render target (back buffer of swap chain or texture) directly. There’s no intermediate render target involved (juhu!). Our test hardware was a recent MacBook Pro.
The node will now also be listed in Stride’s performance profiler (F3 -> F5 in render window).