A company streams IoT sensor data about city traffic into Amazon Kinesis Data Streams. Multiple applications consume from the stream, but several consumers are being throttled and receive ReadProvisionedThroughputExceeded errors. Which actions should the solutions architect take to resolve the throttling? (Choose three.)

Throttling and ReadProvisionedThroughputExceeded errors indicate that consumers are attempting to read data faster than the stream's provisioned capacity. Increasing the number of shards (resharding) directly increases the read throughput capacity of the Kinesis stream, allowing more data to be consumed without throttling. Using consumers with enhanced fan-out provides each consumer with its own dedicated throughput of 2 MB/s per shard, preventing interference between consumers and eliminating shared throughput bottlenecks. Implementing an error-retry strategy with exponential backoff in consumer applications is a best practice for transient errors like throttling. It allows consumers to gracefully handle temporary capacity issues by retrying requests after increasing delays, reducing the immediate load on the stream. Decreasing the number of shards would reduce throughput, worsening the problem. The KPL is for producers, not consumers, and adjusting polling frequency alone might not solve a fundamental throughput limitation. Dynamic partitioning is not a standard Kinesis Data Streams feature for resolving read throttling.