My point is simply that the "overhead" is not material in the vast number of cases. Now, I suppose someone could go and measure two exact same button automations in the two apps (once BC is back) and possibly discover some number of milliseconds of difference (I don't know if they would or not, wouldn't waste my time on this), and from that conclude that "overhead" is a motivator to do something one way or another. I suppose it depends on one's motivations and expectations for a given automations, and how concerned one is about some number of milliseconds.
I'm not dismissing this completely. I have a particular button automation using a Pico that the use context is such that I am very aware of milliseconds in the response. It dawned on me a few days ago that the automation was "slow" because I was using the regular Pico driver, so it was waiting for button release to process button pushed -- and I could see this. So I changed the driver to Fast Pico. This one is in RM with a Z-Wave dimmer. It's virtually instantaneous. Gosh, maybe I should move it to Simple Lighting to make it even more instantaneous.