Risk is a function of both the likelihood and consequences of failure. If you can try and fail a hundred times without quitting your job and only take that plunge when it is derisked (by increased information derived from operation) then it is inherently low risk. Sure throwing 100% into a startup from the beginning makes that individual effort more likely to succeed while being much more risky, but nights/weekend projects can evolve to sustainable businesses without increasing your risk profile. This is because the consequences of failure are near zero, and even a failed effort may be "profitable" in the net sense of connections and marketable skills gained.
Only in a mathematical argument. Failures are difficult to deal with personally, especially if you have to deal with them repeatedly without any sign of success around. The typical result is frustration gradually turning into a form of burnout and depression.
While the consequences you describe can be seen as quite severe, the risk of them is, in my opinion and experience, a lot lower than the risk of failure alone. The chance of failing AND consequently falling into burnout or, less likely, depression, seems rather small to me.
Software is inherently mid- to high-risk. "Low-risk" just doesn't exist in a world where half of all projects fail, and some fail for reasons not due to a lack of merit (but because they were built to a specialized purpose for which there was demand, but the situation changed). So bank loans, which require personal liability are completely out. On the other hand, VCs want to play in the high-risk space where the mean (expected) return is high but the median is failure. That appeals to some people, but it shouldn't be the only game in town.
Right now, there isn't a mainstream funding outlet for the mid-risk businesses oriented toward 10-40% per year growth, even though that's a great space for software engineers, since it approximately matches the acceleration curve of our capability, with hard work.