We have encountered several issues related to the GetRegion call and region cache updates, such as #2843 and #8426. At present, a highly feasible solution to alleviate this type of problem is to change GetRegion from a unary gRPC call to a stream call that supports batch processing, which should reduce the CPU pressure on PD in scenarios such as cache expiration and error retries strom.

• Implement the gRPC stream version of GetRegion which supports handling various queries, such as GetPrevRegion and GetRegionByID.
• Implement the batch processing logic within the PD client.
• Integrate the new GetRegion method into the client-go/TiDB.
• Conduct the benchmark tests with the previous version to evaluate the QPS/CPU usage performance.