There might be some error on your code or inadequate error in the Dataflow job upon encountering insert errors. Issue could arise due to resource constraints on the Dataflow worker, that’s why the job is stopped and the records being dropped, try to check the worker logs for clues. Also, you might want to do an error handling mechanism like retrying inserts or writing failed records to a separate location for you to analyze more, or try implementing retry logic with a dead-letter queue. At the BigQuery side, check the audit logs for the information on insertion errors. BigQuery has limits, try to check the limits related to concurrent insertions or data size per request. When inserting records in batches, a single bad record in a batch could cause the entire batch to fail.