Definitely does not seem to be the case for me in the P85D with .253. Looked out for this a few times today and didn't experience anything that would suggest it monitored brake lights.
Maybe it's better if I describe the behavior and not the conclusion. I'm open to being wrong on both.
With .251, the operational case of decelerating car in front of you, locked on by TACC, is greatly improved. Whereas before, TACC would only decelerate when the distance to the next car was noticeably shortening, it is now doing it the way I would do it as a human. I see brake lights, I let off the go pedal ever so slightly to a coast condition, maybe 0.25-0.5 seconds before old TACC would have waited to decelerate. New TACC seems to do this earlier than before as a coast condition instead of jumping straight to hard decel.
A 0.25-0.5 second difference is huge. Perceptive drivers could describe one version as "choppy" with "roughness" while the reduced-latency version is a non-event in the driver's mind. For those of you who are hyper sensitive to this, you know the difference between 250ms and 500ms lag (both suck!).
Now, for the conclusion bit. I don't agree that when brake lights go on, cars slow down. If you're like me in my former ICE car, you sometimes lightly hover brake as you want to slow down, but you're not quite sure yet. This state could last as little as 0.5 seconds, but usually, a human decision is made by the event +1.0 second mark. In this condition, the brake lights are on, but deceleration is minimal. A perceptive human driver will enter a coast-condition for the initial 0.5 seconds in anticipation, but old TACC did not.
Whether the new TACC is doing some kind of brake-light algorithm on top of the radar, I don't know. But if it were me working on the autopilot team, I would for sure superimpose a brake-light algorithm on top of the radar sensitivity, even if it just gave the TACC a 0.25-0.5 early action in the form of deceleration/coast.
- K