My mailbox was full for two days because I didn't check it for about a week; I've read up on this thread in the archives, but may have missed something from the last few hours. With that in mind, I'll play expert again because I've read the International Code of Zoological Nomenclature... disclaimer: I haven't talked to anyone (such as Henry Gee, any of the authors, or any member of the International Commission on Zoological Nomenclature for example) and have no information other than what is public.
To me it seems crystal-clear that Henry Gee is wrong on the nomenclature, so that Oculudentavis khaungraae remains an available name ("available" meaning "properly published, and not a junior homonym, therefore valid unless deemed a junior synonym which is a matter of taxonomic freedom and not part of nomenclature"). The Code does not know such a thing as retraction even exists, notably not mentioning it in Article 9 ("What does not constitute published work"). Article 8.1.1. says that, to be published, a work "must be issued for the purpose of providing a public and permanent scientific record"; the Code has no concept that this could be true of a work at some point and then no longer true of the same work at another point in time. (Link to all of Art. 8: "What constitutes published work".) In other words, the fact that the retraction has not annihilated all printed copies and all the PDFs on people's harddrives means it is still published. (...I thought there was a minimum number of copies that had to exist, but that's just a Recommendation, 8B to be exact, implicitly also 8A, 8E, 8F and 8H. In other words, if a work published on paper is only extant in one copy in someone's attic, it's still published, and likewise for a work published electronically that has by some miracle vanished from all storage media except one.)
Indeed, deep in Recommendation 8D there is hidden this claim of fact that is not a recommendation but a clarifying restatement: "The content of a work is immutable once it is published." Clearly, this means the content of a work can't be deleted. (The rest of Rec. 8D really is a recommendation, and is therefore worded with "should": "Corrections should be made through notices of errata or other separate publications. Second or other additional printings of a work should be clearly labeled as such, with date of publication stated in the work, even if no changes have been introduced.")
The paper may no longer be available from the journal (I haven't checked if that is so, or if the PDF is still available with a big watermark saying "RETRACTED", which is how at least some journals do their retractions). That, explicitly, does not matter: Art. 8.1.2. says that, to be published, a work "must be obtainable, when first issued [...]" – if it's no longer available now, that's irrelevant.
Art. 8.7 casually "reminds" us (long before we've read Art. 81) that the Commission has the power to decide that a work "is to be treated as not having been published". Nowhere is there any indication that anyone else, for example a publisher, might also have that power.
It may seem, or even be, odd or unfair that publishers can get nomenclatural acts published but not unpublished. But by all evidence that's how the Code sees it, and to me that evidence looks not only uncontradicted but very strong.
The comparison to a Supreme Court case suffers from the fact that US law (except Louisiana state law, and I wonder about Puerto Rico) is case law, while the Code is code law. Principle 8:
There is no "case law" in zoological nomenclature. Problems in nomenclature are decided by applying the Code directly, and never by reference to precedent. If the Commission is called on to make a ruling on a particular case, the decision relates to that case alone.
80.5. Interpretation of Opinions
An Opinion applies only to the particular case before the Commission and is to be rigidly construed; no conclusions other than those expressly specified are to be drawn from it.
In yet other words, Opinions by the Commission are not new legislation. To decide on anything more than the one case a petition would be about, the Commission would need to alter the Code, and the procedure for that is quite different (Art. 90).
On the unrelated matter of inappropriate names, I see your Basilosaurus and raise you Mastodonsaurus (which, on top of that, is a mess language-wise, like Oculudentavis khaungraae). Owen believed he could amend that one to Labyrinthodon. Alas, he was wrong.
On the also unrelated claim that science isn't published in Science or Nature – well, sometimes it is, at least if you count the "supplementary information". Scientists who need to improve their impact factor submit their manuscripts to the glamour mags, if they think the manuscripts are flashy enough; I hope to do so myself later this year. Peer review is no better or worse there than elsewhere (I've reviewed one manuscript for Nature myself, and others for a bunch of other journals – it's not like reviewers have exclusive contracts, or any contracts). The trick is that manuscripts that come to flashy conclusions because the science in them is shoddy are much less likely to be submitted to more modest journals, and that's why the glamour mags may well contain more bad papers than most other journals do: reviewers at other journals simply don't get that many chances to fail in their reviews of flashy papers.
Glamour attracts corruption, and so it's no surprise if Ruben and Bakker got insufficiently rigorous manuscripts into the glamour mags because of personal connections. But even that goes both ways. I know a good paper that was published in a glamour mag even though it can hardly be claimed to be flashy enough for that; rumour has it that it was published there because one of the authors has a personal connection to the editor. I'm honestly happy the paper broke the flash barrier and most likely reached a wider audience than it would otherwise have.