This basically means hostname management not only does not work, but it's very much broken.
Particularly there's no way to assign and keep 2.1.0 node's hostnames during rebalance upgrade path.
Known workaround is doing:
wget --post-data="ns_cluster:do_change_address(\"mysuperhostname.com\", true)." --user=Administrator --password=couchbase http://mysuperhostname:8091/diag/eval
on 2.1.0 node after it's added but before it's rebalanced in.
It is enough to do that only first time during rebalance upgrade. All subsequent nodes can be added via "Add Server" functionality where target server name is used as name to assign.