I have run into another brick wall trying to transform my synchronous long-polling function into an asynchronous one using pipelines. At the end of each request, regardless of outcome, I must schedule some cleanup tasks in case a client doesn't repeat the polling request within a timeout.
Currently I can only see how to handle error and success separately, but no way to provide code that runs unconditionally once the pipeline is realized. Do I have to resort to some kind of duplication, or am I once again overlooking something in the API?