The correct answer is Use a per item random hash as the partition key.

Use a per item random hash as the partition key produces a high cardinality and uniformly distributed key space which helps Cosmos DB spread write throughput across many physical partitions. A synthetic random hash avoids concentrating writes on a small set of partition key values and prevents hot partitions during sustained heavy ingest.

Partition by the review submission timestamp is incorrect because timestamps tend to be sequential or clustered and they can concentrate writes into a small number of partitions during bursts. This pattern leads to hot partitions and uneven throughput consumption.

Cloud Bigtable is incorrect because it is a different database service and not a partition key strategy for Cosmos DB. It does not answer the question about how to choose a partition key to balance write traffic in Cosmos DB.

Partition by product SKU is incorrect because product SKUs are often skewed by popularity and they can concentrate writes on a few hot SKUs. That skew creates uneven partition load and does not provide the uniform distribution needed for heavy ingest.

Full AWS Practitioner Certification Question