Have you tried a max ray depth of 3 (to over compensate for two layers) and upping your voxel scattering quality?
For example, I'm doing an ambient lit scene, no direct light, a single haze layer is noisey as hell, so I had to do voxel scattering quality of 1000, and it's pretty much noiseless even at AA3 for previews. Haze quality was always 1. Though I believe you can exceed maximum quality of a cloud layer. I believe I was using like 5 for my fire.