Problems on retrieving batch when using redshift_datasource

[leodrivera]

I used to use this data souce to connect to a redshift server and it works flawless (Used the redshift with SQLAlchemy 1.4):

    CONNECTION_STRING = (
        "postgresql+psycopg2://${REDSHIFT_USERNAME}:${REDSHIFT_PASSWORD}"
        f"@{REDSHIFT_HOST}:5439/master?sslmode=allow"
    )
    datasource_name = "redshift_datasource"
    datasource = context.data_sources.add_or_update_postgres(
        name=datasource_name, connection_string=CONNECTION_STRING
    )

Then, I've updated to use Redshift GX fork with SQLAlchemy 2.0 with the following data source:

    redshift_connection_details = RedshiftConnectionDetails(
        user="${REDSHIFT_USERNAME}",
        password="${REDSHIFT_PASSWORD}",
        host=REDSHIFT_HOST,
        port=5439,
        database="master",
        sslmode=RedshiftSSLModes.ALLOW,
    )

    datasource_name = "redshift_datasource"
    datasource = context.data_sources.add_or_update_redshift(
        name=datasource_name,
        connection_string=redshift_connection_details,
    )

If I use it and test with a batch:

batch_definition = table_asset.get_batch_definition(batch_name)
batch = batch_definition.get_batch()
print("batch.head():", batch.head())
print("batch.columns():", batch.columns())

It will return the dataframe with batch.head(). However, the batch.columns will return []
This is breaking any expectation I try to test.

Versions:
great-expectations (1.4.4)
gx-sqlalchemy-redshift (0.8.20)