Retry unprocessed keys using exponential backoff with jitter between attempts and Switch to the AWS SDK's batch request client to leverage built-in retries are correct.

Retry unprocessed keys using exponential backoff with jitter between attempts addresses the common causes of partial results such as throttling and hot partitions. DynamoDB returns UnprocessedKeys when the request hits capacity or size limits and retrying only the unprocessed items with exponential backoff and jitter reduces contention and spreads retries over time.

Switch to the AWS SDK's batch request client to leverage built-in retries is also a reliable approach because the SDK implements retries that follow best practices for backoff and jitter. Using the SDK reduces custom code for retry logic and helps ensure consistent, proven retry behavior.

Add a new Global Secondary Index with separate read capacity settings does not change how BatchGetItem returns UnprocessedKeys and will not guarantee complete batch results in the face of size limits or throttling on the primary table.

On failure, reissue the batch request immediately without delay tends to repeat the same throttled traffic and increases the chance of repeated failures rather than resolving the UnprocessedKeys.

Raise the table's read capacity and turn on Auto Scaling can reduce throttling but it does not prevent partial results that occur because of response size limits and it may not solve hot partition effects by itself.

Full AWS Practitioner Certification Question