When you send a prompt to Copilot, there's a filter before that prompt reaches the model, the model inference runs, and then that filters on the output side of the model.
And the hold time is that all this stuff needs to propagate through this logic gate or the flip-flop until the signal really is stable on the output again.