At the moment the clouds are still unanimated and cover an area of 153x153 kilometers (i'm tiling a 16x16 grid of clouds, each voxel space is around 10 km large).
To reduce fillrate, i am now rendering the closer clouds into a small resolution texture (256x256 in the screenshot below) and alpha blend it on top of the screen. I'm not quite sure yet how it will interact with the ZBuffer, but the framerate increase is pretty good, especially when navigating inside the clouds. With that "trick", the framerate is constantly over the 100 fps, and generally around 150 fps, at ground level.
Here is a screenshot taken a bit above the cloud layer: