[William Cloud Hicklin]

Tolkien was driven by a few impulses, often contradictory, and the vector arithmetic depending on small shifts could lead to different outcomes. Tolkien was a complex and inconsistent man!

On the one hand, we have the Patience (Solitaire) player. He favored a form called Forty Thieves (or Napoleon at St Helena), which is a very strategic game, one in which one has to plan several moves ahead as in chess. He also was fond of the Times crossword (which, for non-Brits, is one where every clue is itself a small word-puzzle to be teased out). In short, Tolkien's mind delighted in working out puzzles according to set Rules - even where the Rules were those he had made himself. This is especially the case with his languages, which can be said to be enormous puzzles, modified by aesthetic taste.

But then, as we can see with the languages, he was perfectly capable of changing the Rules if he felt like it. It wasn't Calvinball anarchy, but he still wasn't disposed to make himself a complete prisoner. The Problem of Ros was a voluntary submission to Rules he could change at will.*

Then we have Tolkien the fiddler, the reviser, the niggler, as reflected in Christopher's comment in the OP. He had almost a compulsion to tinker with things, from a little -mo/-vo alteration of the grammar, to the massive uprooting of Noldorin from its elaborate historio-cultural context and its recasting as Sindarin, to the gargantuan destruction and re-constitution of the fundamental cosmological myth, never actually carried out.

Into this add Complexity: everything was so interlaced and interdependent that pulling on a thread, without extreme care, could lead to a large unraveling. This did not mean never to to it, but did incline towards circumspection.

Into this we intrude the apparent, but only apparent, Immovable Object, the vector of inverse magnitude and direction to any intersecting vector: publication. This is really the intrusion of the pragmatic external world: novelists don't get revised editions (not usually). Once it's in type it's too late to change anything. But this isn't to say that Tolkien didn't chafe under that externally-imposed Rule, and wish he could change some printed material anyway: this is exemplified by the "sample" Riddles in the Dark which he sent to A&U on a wing and a prayer and which, much to his amazement, was incorporated in a new edition. And the same applies to the 2d Ed. LR: he wasn't expecting it, but when the opportunity presented itself, well, then shackles were loosed and he felt free to implement changes he had wanted to make. Among these changes were the names of Finwe's third son, and his son.

Add to this the fact that Tolkien was a very imperfect editor, and often missed things he should have revised to fit changes but overlooked Hammond and Scull are full of these, such as Aragorn's "three nights" which escaped correction to "four" in The Riders of Rohan; as well as the Company's "previous" night in the flets which was no longer "previous" after the night spent on the ground had been inserted. (Many of these were caught later by Christopher, still paying attention to Dwarf hood colors. CT was rather more meticulous than his Da.)

And so I don't think publication forms the Iron Curtain of 'canonicity': it may, or it may not, represent Tolkien's ultimate intention; what it represents is an externally-imposed pragmatic limit, which Tolkien sometimes, but not always, believed he had to surrender to, as with -ros. He had got a 2nd edition, but had no reason to expect a 3rd.

________________________
*I have long felt that he could have worked his way around it by postulating that vernacular Gondorian Sindarin wasn't the pure book-form, but contained a gaggle of loan-words collected over the millennia from Adunaic, pre-Numenorean local languages, Rhovanic/Rohirric and probably even the Haradrim. Just like the vernacular Latins that became French,** Spanish and Italian. The element -ros was a surviving loanword from Beorian via Adunaic into demotic Sindarin, preserved probably due to the name of the first King- and probably assumed by the nonlearned to have always been Sindarin.

**Roughly 10% even of modern French vocabulary derives from Frankish