Having both worked in startup/tech companies and ad agencies, one of the most valuable skills I've learned is being able to decipher and debug someone else's spaghetti code. Clients shift agencies all the time and sometimes you're just given the entire codebase for site, with no documentation and no help offered from the previous agency, and you have to make an update to it or learn how some vital part of it works so you can recreate it in the new site. That's happened to me quite a few times and it made me a better debugger and coder.

There is a difference between writing broken code many times while never learning, and writing broken code once learning why it was broken then starting over and writing good code.

If you are saying you can (and should) fix code but cannot fix, for example, a joke. I disagree.

Only once you know what the problem is can you decide if it makes more sense to start over.

I have no idea whether you can fix a joke. I'm not very funny.

I disagree.

I somewhat disagree. I'm in the process of building my 5th production rails app, and 8th web app (which will be used in production and make a profit). I've gotten ridiculously better at figuring out how to structure the underlying models.

This alone has made me way better at structuring databases for queries, planning properly, etc. My day job is more of a data science role and it's helped a ton to know how to build the whole stack.

I think a closer analogy is fast iterations vs waterfall design.