There is a powerful mechanism called "backpressure" derived from mpsc::channel. Alongside the function of data transfer, we could also utilize the channel with this mechanism to achieve a specific effect.

"Backpressure" relies on the async runtime. When a bounded channel is full, the sender will suspend the current task and yield control back to the async executor until it becomes ready again. This only works in an async context. Otherwise, the sender will just block. How can we use the phenomenon? In my experience, this can be used to automatically "sleep" background tasks. In most cases, the data flows in one direction. Once we control the consumer at the endpoint, the tasks along the entire pipeline gracefully "sleep." This can save a lot of work if we want to temporarily stop the consuming process and the upstream tasks. Symmetrically, it works in the scenario that we make the producer side stop sending.

Insight: There are many elegant optimizations leveraging the side effects from the original action. In addition to "backpressure," I have encountered mechanisms like storing information in the unused bits of a chunk of memory, storing static assets in the static space of an executable file etc.