> But wouldn't salaries for software workers eventually rise?
No.
Lawyers, accountants, and teachers can make a lot of money because people are willing to pay a lot of money for the services they provide, because those services are very valuable.
Paying a top-notch lawyer US$500 an hour can save you a huge multiple of that US$500 if you're being sued for US$500M. Paying a good accountant allows you to manage your business transparently enough that you can raise money on a stock market, thus getting 20 times the amount of cash you've made in profits. A four-year university education can lift you from the lower class into the middle class, adding millions to your earnings and a decade to your lifespan.
It turns out that developing software quickly is a very valuable activity. In a single day of work, a good programmer can write a piece of software that functions in production for years afterwards, serving dozens, hundreds, or millions of people. Often, this is even the case if the software is buggy; the alternatives (repetitive manual labor, spreadsheets built by amateurs) are often even less reliable.
If you tried to outlaw the development of unreliable software, programmer salaries would go the same place that salaries for nuclear plant engineers and designers of small planes have gone over the last three decades: overseas or into oblivion.
Worse, it would likely be counterproductive. What we need for more reliable software is not books of management rules but better abstractions, better theorem provers, better programming languages, better insights. Those things are themselves software, but they are probably not software that we can develop faster, or at all, under an SEI CMM Level 5 process.