Well, yes, agreed - that is the essential domain complexity.

But my argument is that we can work to minimize the time we spend on verifying the code-level accidental complexity.

Sure, but that is what we've been doing since the early 2000s (e.g. aslr, read only stacks, static analysis, etc).

And we've had some succeses, but i wouldn't expect any game changing breakthroughs any time soon.