Sagemaker Feature Store Spark connector is duplicating data

Hi!

I'm using the Feature Store Spark connector to ingest data into the Sagemaker Feature Store and when we try to ingest data to a Feature Group with the online store enabled, the data is duplicated. In the image bellow, "customer_id" is the ID feature, "date_ref" the event column. All the features are equal for the same ID and EventTime column, except the "api_invocation_time".

If the feature group doesn't have the online store enabled, we ingest the data directly to the offline store without issues. But when we use the "Ingest by default" option in the connector (not specifying the "target_stores" in the connector, uses the PutRecord API), the data ingested is duplicated:

params = {
    "input_data_frame":dataframe,
    "feature_group_arn": feature_group_arn            
}

if not online_store_enabled:
    params["target_stores"] = ["OfflineStore"]
    logger.info(f"Ingesting data to the offline store")

pyspark_connector.ingest_data(**params)
logger.info("Finished the ingestion!")

failed_records = pyspark_connector.get_failed_stream_ingestion_data_frame()

How can I solve this issue using the connector?

EDIT:

Apparently, the problem is in the "get_failed_stream ingestion data frame" method. This method, instead of just returning a dataframe, ingests the data again before returning the failed records. Removing the method from the ingestion pipeline resolves the issue, although we lose a form of validation.