The AsyncRequestsSender doesn't check if an incoming request is intended for the current node. In this scenario, the ARS will schedule a remote command instead of running it locally. This introduces unnecessary overhead, as the operation must acquire a connection, transmit the command, pass through the full ingress and command dispatch pipeline, and then return a response over the network.

We should investigate updating the AsyncRequestsSender to run a command locally if it's intended for the current node.