>* ... legally enforced ...*
Other than that part (most countries in the world do not have regulations or licensing requirements for most engineering disciplines) I would agree. But I would also point out the set of software projects that meet that definition is much larger than those you listed.
As mentioned, it's a matter of economics, so the rigor scales with the pain it can cause if something that goes wrong. Hence any software that has a high blast radius is that rigorously built, probably even more. There are entire categories (not just individual examples!) of such projects. An obvious category are platforms that run or build other applications: OS kernels, databases, compilers, frameworks, cloud platforms (yes those 9's are an industry standard), and so on.
Then there are those regulated ones like automotive, aviation and medical software. There is even a case to be made for critical financial software.
Another less obvious category applies to any large software services company that has oncall engineers, because the high cost of engineers quickly climbs and quality processes quickly get installed, which basically amount to those critera you listed.
That internal LoB app with 5 users? That level of rigor simply does not make economic sense. Which is probably what you mean by:
> 99.9% of the time though, software "engineering" is an ad hoc, mix and match, semi-random, always changing requirements and environments, half-art half-guess, process, by unlicensed practicioners, that is only regulated at some minor aspects of its operation (like GDPR, or accessibility requirements), if that.
To that I'll say, as someone whose first site outage as an intern was an actual industrial manufacturing factory (not an AbstractFactoryFactory!) a surprisingly large fraction of projects in other engineering disciplines match that description ;-)