We have been looking at it already, some time ago. The problem was that it was all CPU-generated (no wonder Intel picked it to support), not ready for spherical worlds, and mainly problematic to integrate with the existing atmospheric scattering code and lighting. There are some other options, but I fear that for the best performance and seamless integration we'll have to develop it.
The domain is well suited for parallel processing on the GPU, and we can also employ our procedural techniques to refine the detail and make it play nicely with the rest of the system.