Even in a slow computing initiative, outside invention will continue. Take ASLR, DEP, and other memory corruption mitigations. The architects of the legacy, fixed-address systems didn't expect the large amount of vulnerability due to buffer overflows in programs that handle external input. The authors of these programs failed to write tests that exercised some edge cases of malformed input. While long/malformed input unit tests might seem germane today, they weren't then.
Custom software initiatives make sense for people who have a high aptitude and the time to undertake them. This use case doesn't fit the vast majority of people who use computers.
While Microsoft certainly deserved the $10k judgement against them, software authors have the duty to help mitigate past imperfections, errors, and lack of foresight in their software. Authors of free and proprietary software share this responsibility.