I'm a bit more awake right now, so I can elaborate a bit more on what I make of this.
What I understand about the run:
1. It's a slower run than the current published run.
2. It can fully complete the game from start to finish from any starting state on any unmodded cartridge.
3. It forgoes using timesavers that are currently impossible for a bot to use in real-time (in this case, resets)
Based on (1), I don't think this run should obsolete any runs of this game (but that was obvious... right?)
Normally, the deliberation would end there, but because of (2) this run is unique in that is possesses a feature that the published run doesn't- it can be replayed on a console.
People who aren't part of the TASing community don't really understand the mechanics behind these god-like runs they are seeing, and I'm sure all of us have groaned in annoyance whenever we read a comment or hear someone say that a TAS is "cheating!" Seeing the run accomplished a real console shows that such superplay is possible without mods, and makes the community more accessible to outsiders.
But maybe you really hate new people, and think the TASes published on this site should cater only to members of the community. You argue that the only runs accepted should be as fast as possible, and the best tool-assisted speedrun should be the best run in ideal circumstances. The folly in the argument is that even when using an emulator, ideal circumstances cannot be achieved. (compared to a console verified TAS, where ideal circumstances are nigh-impossible to achieve).
A year or two back on this site, there was a discussion about some games that displayed different cutscenes/background gameplay during the start screen, but the emulator would always start with the same cutscene/background. I think it was around the time the acronym TAS got changed from "Tool-Assisted Speedrun" to "Tool-Assisted Superplay."
Anyway, where I'm going with this is that even with an emulator it is impossible to achieve ideal circumstances for a flawless run-through of a game (and choosing an ideal startup state would violate another (somewhat arbitrary) rule). Ultimately, since you already cannot achieve an ideal run, saying that this console-verifiable run has less merit is a hypocrisy.
Another lesser argument for the publication of the movie is that there is an impressive technical achievement in creating a run that can be replayed from any start-up (even from an older save-file from a ROM).
Based on these points, I think (3) is a similar, weaker argument akin to (2).
I enjoyed the movie, I think it's a breakthrough in console verification and a milestone in showing that even the most complex RNGs can be overcome. I believe the movie should be published based on these merits.
Yes vote.