Make up your mind! If the level add guy and the level subtract guy had just talked to each other they could've just left the levels as they were and saved everybody a lot of work.
I think what's actually happening here is that Mario World came from a lineage of arcade games, where you start from the beginning every time you play, but was pointing to way towards the style of game where your progress is saved and you start each session from where you left off on the previous one, which was a more natural fit for longer games.
Warp zones make sense in an arcade context because you start each session from the beginning of the game because advanced players might want to skip to the interesting levels. I'm sure it felt natural to bring that kind of secret forward into sequels. But in Mario World your progress is saved between sessions, which creates a new kind of interaction where you exhaust a game's content rather than replaying it. In that situation, new content to exhaust becomes a more natural reward.
we don't think the differing pressures between arcade and home games were quite clear to everyone at the time. this was the first main-line home-console Mario game to have a way to save progress, lots of things were experimental.
it's also interesting to compare how you look for secrets in SMW with games before and after it. in the NES era, there was this idea - widespread across all home games, not just first-party titles - that you should be exhaustively checking every tile in the world to see whether it had something hidden in it, and then memorizing the locations of the secrets you found so you could find them again next playthrough.
this was fun. we say that because younger audiences might not realize. :D
(personally we still have trouble getting into any game that doesn't do it. alas, that's every game these days.)
as there started to be more games in the world, and ways to save progress other than knowledge checkpoints, this stopped being fashionable. in a world where several hundred games you might want to play come out every single year, you aren't likely to want to spend most of the year exhausting the potential of any single one, that's just arithmetic. plus the existence of passwords and battery RAM made it kind-of obsolete.
of course, it's also the case that the science of user experience has advanced significantly in the intervening decades. we know how to cue players towards finding "secrets" in ways that feel satisfying, now. (at least, they feel satisfying to people who didn't grow up having to do exhaustive search, and who don't understand the game design principles behind the cues. if you can read every nudge the game is trying to be gentle about as if it's a direct dialogue with the designers, that's a lot less satisfying. oh well. we made a bargain for knowledge and this is the price. someday there will be none like us left, and then it will be fine.)
but..... this is something we're able to identify today, looking back. at the time, it had to be lived. it's seldom obvious exactly how the world is going to change, while it's happening; if it were, the change would have already finished.
so when we're talking about historical game design practices, remember! some of this stuff was experimental and people didn't know exactly how it was going to play out. the general public knows more, today, about the implications of the design decisions in Super Mario World, than its designers did at the time.