Issues can be reported on UDN if you have access otherwise here.
Cloud shadow map: that has been only designed the top of the planet also.
Weird blue color on cloud: when we sample the aerial perspective on cloud, we only do that once instead of per sample for the sake of performance. It is done using a single distance according to each samples and their coverage. It can cause such visual discrepancy, but it is the first time I see it that strong. We could add an option to evaluate aerial perspective per sample but that is going to cost more.
I take notes of those issues too.