A. With Windows Server 2008 R2 Hyper-V CSVs, all nodes in the cluster can directly write to the CSV volume except for certain types of operations. File copy operations are an exception, because they're expanding write operations, which change metadata, and therefore have to be redirected to the coordinator node. The coordinator node is the cluster member that has the CSV volume locally mounted and can therefore make metadata changes.
This expanding write is redirected over the cluster network. In a VHD copy operation, there are a lot of these expanding write operations that, if not performed on the coordinator node, will result in a lot of redirected traffic. So the best practice is to always perform large file copy operations to a CSV on the coordinator node for that CSV volume that's being copied to. The good news is that VMM is optimized for CSVs and will perform the copy on the coordinator for the target CSV volume for file copy operations.