The data engineering team at an e-commerce company processes transactions into Amazon Kinesis Data Streams using the Kinesis Producer Library (KPL). The Data Streams are managed via Auto Scaling configuration. On the other hand, the Kinesis Client Library (KCL) ingests the incoming data into the company's warehousing system to be used for downstream analytics. Lately, the data engineering team has come across issues arising out of duplicate records. Which of the following would you identify as the most likely reason for this behavior?