1 - Im no gluster expert but - moving blocks on to a high performance machine could be really good for long running mapreduce jobs wherein there are large amounts of slow disks in a cluster.
2 - I assume you would rather discourage people from custom coding the brick logic - since its the volume level of abstraction that you want people normally to work from - right... ?
3 - Are there optimizations that happen in the way the gluster fuse mounts work, wherein volumes sort of assume that the bricks aren't moving around beneath them.?