Attention signal delay during complex join execution in SQL Server

Request Cancellation Behavior

Microsoft SQL Server utilizes the Attention signal to notify the execution engine when a client-side command timeout expires or a user manually cancels a request. While this mechanism is designed to stop batch processing, the server must reach a safe checkpoint within the execution plan before the operation actually terminates.

Resource Contention and Latency

In scenarios involving high-resource operations, such as large CROSS JOINs or complex asynchronous processing, there is a known gap between the driver sending the Attention packet and the server releasing held locks. This creates a state where the client perceives a timeout, but the server continues to consume CPU and memory resources.

• Version Assumption: SQL Server 2019 or later.
• Verification Method: Monitoring sys.dm_exec_requests during a manual cancellation via SSMS.

What determines the specific checkpoints where the execution engine acknowledges the Attention signal? Is there a configuration to reduce the latency between the signal receipt and the actual termination of the process?