Java String Exception thrown when exporting data

My graph export code has been working well, until now a new node property 'graphEmbedding' of 256 dimension is added to each node. The relevant export code is below:

Export node ...
MATCH (m:Product)
WITH collect(DISTINCT m) AS M
CALL apoc.export.csv.data( M, [], null,  {stream:true, batchSize:2000000}) YIELD data
WITH apoc.text.replace(data, '(?:,"_start","_end","_type"|,,,)(?=\n)', '') AS mdata
RETURN mdata

The exception is:

File "/Users/congminmin/.venv/wbkg/lib/python3.8/site-packages/neo4j/work/result.py", line 227, in __iter__
    self._connection.fetch_message()  # Receive at least one message from the server, if available.
  File "/Users/congminmin/.venv/wbkg/lib/python3.8/site-packages/neo4j/work/result.py", line 62, in inner
    func(*args, **kwargs)
  File "/Users/congminmin/.venv/wbkg/lib/python3.8/site-packages/neo4j/io/_bolt4.py", line 271, in fetch_message
    response.on_failure(summary_metadata or {})
  File "/Users/congminmin/.venv/wbkg/lib/python3.8/site-packages/neo4j/io/_common.py", line 183, in on_failure
    raise Neo4jError.hydrate(**metadata)
neo4j.exceptions.DatabaseError: {code: Neo.DatabaseError.General.UnknownError} {message: This encoder depends on java.lang.StringCoding, which failed to load or apply: Index 16384 out of bounds for length 16384}

I am on MacPro and use Neo4j 4.2.6. 4.0.4 has the same issue. Because the graphEmbedding is a float vector of length 256, probably that's the reason causing the error: " Index 16384 out of bounds for length 16384"

Once I remove the graphEmbedding property from the graph, the export code runs well.

I used the method introduced here to create the embedding:

I registered an issue here: