Nach, I have a question for you. I think it fits better in this topic, as it is based on the SMB judgment. The question is what you think of the following. It's based on three quotes, two of which are yours. One important assumption is that the PAL SMB submission is, in every relevant way, faster than the NTSC published movie.
1) In this topic, you say that "If this was regarding Super Adventure Island or Lemmings, we probably wouldn't be having this conversation at all."
2) In your
decision tree - which is very clear, so thanks for that - the relevant part is that "having a non-original replace a perfectly valid original seemed lunacy to me".
3) The
judge guidelines - Managing game versions/ports on multiple platforms - states that one version is preferred, based on originality (or, age of publication), popularity, or superiority.
I think it is clear that the preferred version for SMB is now based on the popularity of the American version, which is perfectly valid. That is why Lemmings is not relevant, because popularity would not be considered a good basis for choice of version. Since there are few games as popular as SMB, there are few games that are potentially relevant, and even less games that have a version difference. One that may come close is The Legend of Zelda: The Ocarina of Time. In its publication history on this site, the version used has changed from JP 1.1 to USA 1.0 to JP 1.0. Surely this is because of the glitch-potential (so superiority) instead of originality or popularity, even if it is a popular game. So there exists a precedent to choose something else than popularity, even for a popular game, as the basis for the version to publish. Therefore I think that the judge guidelines, as the OoT example shows, provide opportunity to at least soften the "lunacy" of obsoleting a "perfectly valid original".
Furthermore, as MrWint explained, PAL SMB is not just a port, but also a sort of 1.1 patch applied at the same time as it was patched for 50Hz screens. The reason given to prefer originality is authenticity. I do not think it needs explaining that a later revision by the creator is no less authentic than the so called original: after all, the original only happens to be (coincidentally, as it were) published first. In question form: what is more authentic, a modern game with or without its day one patch? Therefore I think the preference of originality (which American SMB has over PAL SMB) is problematic, as the whole idea of preferring the first-published is an arbitrary preference.
I think the guidelines do not dictate a rejection of PAL SMB, nor do they require an obsoletion of NTSC SMB, as they say: "we generally preferred one version of a game." To adhere to this preference for one version only, and have this version be based on its popularity is therefore a valid judgment, but not necessarily the only valid judgment.