Engine braking with an internal combustion engine generates a vacuum in the intake manifold, because the throttle is closed while the pistons suck air.

(Performance cars sometimes have vacuum gauges to measure this. Aftermarket ones can be installed.)

A brake light could be rigged to activate past a certain a vacuum threshold. (There would be some false positives that are possibly not worth caring about.)

For all the engineering described in the article, I'm surprised it doesn't mention this possibility, if only to give reasons why it was discarded. (Maybe it's a bad solution; I have no ide!)

It seems that a vacuum-driven brake light could possibly have an advantage of kicking in faster than a motion detector, because it could trigger as soon as the revs are dropping with the throttle closed, before the clutch engages to actually connect the engine braking to the wheel.

I.e. blip-throttle before downshift -> vacuum kicking in / light comes on -> downshift completes, actual braking begins.