The two computers are developed by two independent teams who are not allowed to talk to each other. Two different CPUs, two different algorithms, two different programming languages.
The idea, of course, is a design problem with one does not propagate to the other.
If one program is used in all your infrastructure, all your infrastructure is compromised when that program has a bug.
As a marine saying goes, never take two chronometers with you to the sea. Take either one or three.
1) Build servers: Maybe by keeping it possible to build one's software on a laptop — then, when it's time to release something, one can build the software at the build server plus some people's laptops and compare hashes.
2) ???
3) Different people who don't talk with each other, do security code review, independently of each other? Hmm does that make sense
And we should have checks built into the build pipelines and checksum the files that go into the build. We have Merkle trees for a long time. I think, it would be possible to detect injected code.
I think the elephant in the room is actually a more general issue that is cross platform and independent of the product implementation.
It’s 2021 and we are still ignoring the fundamental “best practice” that we’ve known about for at least 20 years.
Systems should be isolated from each other unless there is an overwhelming need for them to be connected and everything should run with the least privilege.
Centralised credential vaults and binary artefact repositories are an obvious example.