An in-depth discussion has been going on over at the neo4j google group discussing the merits and issues related to the explicit node identity of the REST API versus Cypher access.
Michael Hunger and Andreas Kollegger from Neo Technology both participated in the discussion:
We'd like to make Cypher the only thing needed for working with the graph. We've been discussing the difference between identity and addressability, with an underlying assumption that Cypher is the primary interface for working with graph data. Either a node has a unique identity within a label, or has an address (unique pattern match from an identifiable node). The problem with addressing is that nodes that look the same still need a unique local feature to make it easy to work with them.
-- Andreas Kolleger
It's a good technical conversation for neo4j users and people looking into neo4j that want to get a better idea of this particular area.