Currently, once the reasoner responses after a tool call, the conversation assumes that it's the result of the tool call. However, sometimes the reasoner is just giving task updates, which is the case for most long tasks. The conversation needs to understand that it's not function call output and it should continue to wait.