A company has an application that uses an Amazon DynamoDB table. A solutions architect finds that many requests to the table are not returning the latest data. Users have not reported other database performance issues and latency is acceptable. Which design change should the solutions architect recommend?

The problem states that requests are not returning the latest data, indicating a data consistency issue. DynamoDB offers two read consistency models: eventually consistent and strongly consistent. Eventually consistent reads (the default) maximize throughput and minimize latency but might not reflect the results of a recently completed write. This aligns with the observed problem. Strongly consistent reads ensure that all successful writes are reflected in subsequent reads. While they might have slightly higher latency and lower throughput compared to eventually consistent reads, they guarantee that the latest data is returned, directly addressing the problem. Adding read replicas is not applicable to DynamoDB tables for read scaling in the same way it is for relational databases; DynamoDB handles scaling automatically. Global secondary indexes are used to query non-primary key attributes efficiently, not to solve read consistency issues for the main table. Requesting eventually consistent reads would exacerbate the problem, as it's the default behavior that is causing the issue.