The important thing is that identical inputs yield identical outputs. The rest of the vehicle doesn't care what's in the black box.
Then you wouldn't have to warehouse old, obsolete parts. You just have to warehouse old, obsolete embedded-software developers. That's probably cheaper if you have enough volume.
Also, if your company is still programming in assembly instead of a higher-level language, such as C, which has been available for 45 years now, I can safely say you are all hoist on your own petard. If you really want reliability, there's always Ada. You can't expect the 80196 to stick around forever.
The vehicle does not care what is in the black box, but that is only true once input/outputs are actually identical. It is .01 volts different has killed us in random failures. It is 1ms faster, is that better or now - in some cases that 1ms slower was depended on and we didn't know it anymore.
We don't program anything new in assembly. However for a lot of old embedded controllers if it works don't break it. There are no known bugs with the old code. Suddenly the CPU isn't available and we have to create a new one. Marketing can't come up with any new feature that a new CPU would enable so it is strictly and engineering exercise.
If you don't maintain it, you are allowing it to break through neglect. That's like saying if the old oil works in an engine, you shouldn't bother changing it. It works great, right up until it doesn't, and then you have a bigger problem.
Software development is an incurable disease. Once you start, you can't ever stop.
We have a lot of embedded systems written in assembly for an 8-bit CPU that is no longer made. The electronics are easy (I'm not an EE) but it will be years before we can replace the software that currently has been running in the field with zero known bugs - I'd love to get those CPUs that work perfectly made again. Instead we bought a warehouse worth, and we hope it is enough to satisfy all our warranty needs.