Hi Andreas,
the 2 billion records limit also applies in this case, as the records need to pass the general processing of materializing records and inserting them into the table - even though it's the same table.
The distribution onto the different partitions has to happen after that, which means the 2 bln. records limit must be obeyed and cannot be bypassed e.g. by running "partition-local" INSERTs.
As you might agree, the self-insert is a rather seldom real-life use case which thus doesn't have a special optimization.
So, in order to copy your data, you may do it in batches of max. 2 billion records each.
- Lars