Issues:

We should not be recomputing clob hashes.
The lob chunk logic is not guaranteed to fill the buffer.
The CharSequence logic for non-random access lobs is inefficient