A company has two systems on Amazon Kinesis Data Streams: an inventory management system that uses the Kinesis Producer Library (KPL) to publish, and an inventory reordering system that uses the Kinesis Client Library (KCL) to consume. The stream autoscaling is enabled. Before production deployment, the reordering system received duplicate records. Which factors could have caused the duplicates? (Choose two.)

Duplicate records in Kinesis Data Streams can occur due to retries. If the KPL producer experiences network-related timeouts, it might retry sending records, leading to duplicates if the initial send eventually succeeded but the acknowledgment was lost. Similarly, changes in the number of shards or KCL record processors can cause duplicates. When a KCL application rebalances shards among consumers, a record processor might restart processing from an earlier point in the stream, leading to reprocessing and duplication. The IteratorAgeMilliseconds metric being high indicates consumer lag, not a direct cause of duplicates. AggregationEnabled (KPL feature) and maxrecords (KCL feature) do not inherently cause duplicates; aggregation combines records, and maxrecords controls batch size.