Maybe I'm just biased, but I believe in using the simplest tech that will do the job. You don't need an Arduino to blink an LED.
Same as a throttle. That used to be a wire (perhaps it commonly still is?) but the throttle input shifted from being solely the driver into being other systems like cruise control. As soon as you have an electromechanical system it does seem easiest to keep it electric->mechanical rather than mechanical<->electric. That simplification does ignore the driver feedback, however.
Throttle-by-wire is a different beast entirely. If you step on the gas and don't go, that's usually very annoying, and depending on circumstances could be dangerous. But if you step on the brakes and don't stop, that's pretty much a guaranteed wreck right there.
That being said, there's a time and a place for software. Software is a solution to a problem, and I'm not clear on what problem this system is solving. Sure, Bosch's product makes the brake system smaller and lighter. Cool. To what extent does that make the car smaller and lighter? Amdahl's law applies to more than computers. Are we really optimizing the right thing here?
To be sure, when measured against prevailing market trends, this is objectively not a good thing. For safety-critical systems like a vehicle's brakes, I fully expect much more than the superficial considerations I would normally give to appliances.
A mechanical linkage of cables or rods or a hydraulic system all have their failure points also, but the differences are
the infrastructure required to manufacture a stick or a tube vs a chip and solenoids, motors, sensors,
the fact that a cable or linkage or hydraulic piston usually fails gradually over a long time while still functioning, while a chip dies without warning and completely,
and is not repairable in the field except by replacement, while you can repair a mechanical device essentially indefinitely with essentially any tools and materials that are available.
Everything goes to hell if you need to connect to the internet or use high-level software.
The objective measure of what qualifies as "well designed embedded systems" is inherently gray; and where industries have seen some semblance of consensus as a matter of unconditional need, it has always been accompanied with the tears of executives complaining of non-trivial expense.
Isolated embedded systems can be like mechanical systems in that they're simple enough to be tractable and even deterministic.
A system running Linux and calling AWS lamda functions through the cell network isn't tractable.