It's a Starship. It's still in development. It's not a finished product like Falcon. And it's not an unexpected outcome either — after all, SpaceX is doing something that no one has done before, so there does not exist any prior knowledge about the behavior of rockets this huge, and especially reusable. They aren't failing, they are making this knowledge so they could build a rocket that does not explode.

But then again, comparing rockets to software is unfair. Rockets have a finite scope. They go up to safely put things or people into space. In case of SpaceX, they also preferably come back down in one piece to be reused. The more specific requirements only change as a response to new discoveries in the development and testing process — not because some manager has nothing to do, or infinite exponential growth needs to be shown, or investors are demanding AI to be shoehorned into every product, or some designer is desperate for promotion.

> no changes were made, but now the program doesn't work

Some changes for sure were made, because otherwise that would violate the core principle of computer science that the same algorithm executed with the same inputs will always yield the same exact result.