Very often people just take the shortest path from a to b, every single time. So you start with a reasonably shoddy prototype, but then you add some small feature, repeat 1000 times and now you still have a shoddy prototype but it's actually a pretty big project and it's all completely cursed because at no point did anyone do any actual software engineering. And of course now it's too big to rewrite or fix so the only way forward is to keep building on this completely broken mess.

At some point you need to scrap the prototype and replace it with something proper, or at least have a solid plan for how you're going to morph the prototype into something proper. This is often challenging and time consuming work, so a lot of developers tend to never really do it. They just execute the shortest path to get each new feature implemented over and over for years while number of bugs keeps increasing and velocity keeps decreasing because nothing makes sense, everything is more difficult than it should be etc.