Timestamps were chosen over vector clocks intentionally.

"...the primary use case for classic vector clocks of merging non-conflicting updates to different fields w/in a value is already handled by cassandra breaking a row into columns."

See https://issues.apache.org/jira/browse/CASSANDRA-580

If I remember correctly from the talk (I was one of the lucky attendees), Rick claimed basically it is not an issue. Standard time synchronization methods are sufficient for the problem domains Cassandra is meant to address.