When I first played this game years ago, it seems like the kind of game which should be easy to break. I'm very satisfied that it happened, though it wasn't discovered by me. It seems that a few days ago, folks in the RTA community discovered a use for an OOB glitch which was languishing in obscurity. The ability to OOB using frame buffering was discovered sometime in 2014, before the current published TAS (though it seems no one here knew about it). However, this run uses an RTA trick that abuses incorrect collision on one corner of the map.
Once outside, you can walk to this door, and cast Flippendo. This moves your hitbox slightly, and I think touches the warp trigger. However, because you are still in the tutorial, the next "cutscene" hasn't been set yet, and is still initialized to the ending credits (thanks for the clarification on this from StarrlightSims). Thus, this is a Wrong Warp glitch.
Because this is lowering the runtime from ~1hr to under 3 minutes, this should probably be an new Category (Game End Glitch). I hope you are all as satisfied seeing this game destroyed as I was.
Most of the movement is taken from the 1 hour any% run, though I did optimize it a little bit further. You can buffer a frame of input prior to starting dialog, and this includes the button that ends the dialog. So every text box closes 1 frame faster in this run.
feos: Added DaJaWi to authors.
feos: The way the game behaves after this ending is identical to to what it does after a normal ending, and also the explanation of why it warps you to game end provided in the thread looks legitimate to me.
However there are problems with optimization in this movie. It takes vast majority of input from the existing publication and only adds a corner clip and a few seconds of walking out-of-bounds. But even that little part can be improved as shown in the thread, and also the star is not necessary anymore, but the old input wasn't fixed to remove star collection.
We can safely call this submission sloppy, and in normal conditions it would definitely be rejected. But there's a specific clause that applies to Vault movies that says:
In rare cases, there will be a sloppier movie that is faster than a more optimized movie due to the use of a major skip discovery. In this case, the faster of the two movies is preferred.
Last time this happened to me, when I was able to easily beat the submitted times, I talked to both adelikat and Nach about the above clause and asked if it's actually what we intend for the site, even for sloppy movies. And the definitive answer was "Yes": Vault's main goal is hosting records, and the record value is more important in such cases. If it's sloppy, people may improve it later if they care.
I discussed it once again with other judges and the agreement seems to be that it's not full of glaring mistakes that lose a lot of time, the known improvement that's still possible is less than 10 frames so far.
But the clause that may make this movie acceptable is only for improvements! And this run is drastically different from the existing publication, that is in Moons. Well, the problem is that its rating is quite low after 6 years. It has had the exposure that Vault movies don't have, and yet it hasn't really entertained people: there's only one 6 for entertainment, and all the rest are lower. Despite of having gotten okay feedback when it was submitted, publication ratings still outweigh submission votes.
It may be seen as a problem that people who vote in threads aren't interested in rating the actual movies afterwards, and a part of that problem is that the entire rating system is not nice enough to encourage people to mass-rate things. But that doesn't mean we should assume that the overall rating for any given movie would be better if simply more people voted. There'd still be normal distribution and stuff like that: more people rating movies includes more of those who like it and more of those who dislike it.
So far though, we should work with what we have. And the current data we have means the existing publication belongs to Vault. So is this movie, if it's accepted. Which means the Vault clause on sloppy (yet major) improvements neatly applies to this situation, #6451: MESHUGGAH & AIVV73's NES The California Raisins: The Grape Escape in 04:15.83 being an exact precedent (or at least one that I remember).
Accepting to Vault as an improvement to  GBA Harry Potter and the Sorcerer's Stone by DaJaWi in 56:17.96.
feos: This may sound like I'm deliberately making annoying decisions to annoy people into changing the situation those decisions were caused by. Yet it looks like the Vault clause this run was accepted under, is annoying enough for people to go out of their way and improve the hell out of such a questionable movie! This run has been bested in this new submission that looks great! Rejecting.