My experience is in a big-ish software company producing one huge fully integrated bit of software (for phones) probably has some small similarity with avionics in the sense of having to be some combination of hard and soft real-time, handle a lot of different sensors and communication devices and then present a pretty display to humans. Perhaps one a major exception would be not using formal methods in any way (not that I know if avionics software does that).
Like any project, people start out with some design goals (e.g. "it must be hard-realtime" or "every process must handle out-of-memory situations gracefully" or "must run on cheap hardware with one less ram chip so we can make $$$").
These things, especially the ones described by short sentences, tend to have absolutely incredible impact on the difficulty of designing the whole system. They all sound like good things, of course and it's hard to argue against them.
Once you've set out on some paths, it is exceedingly difficult to turn back even if you find out that your initial idea was only partially valid or ceases to be important over time. Every "brick" is laid upon thousands of others.
This is why you need competition and people who rethink from scratch - throwing away some of the assumptions that others have made and seeing what can be done without them either in the light of experience or as a total "what-if". Big companies don't decide to do things like this - they are always trying to have "only one" of everything for efficiency which makes them sterile environments.
I get the impression that cars are becoming more like phones and moving from being all-microprocessors driven by assembly language to more complicated systems with all of that plus big general purpose computers that run user interface and provide conveniences for the driver. It's rather like the nervous system of an animal that's suddenly getting a large brain.
Perhaps aircraft were or are becoming like this too and if so then I imagine that there are few limits to the incredible things that are possible. I think it would be helpful for people who can at least fly and possibly fly fast to be embedded in the software teams however since it is very hard for programmers to make effective trade-offs without really understanding - they tend to get the wrong impression and make the wrong kinds of tradeoffs.