[mhagain]

Thank you. You make a very compelling argument against, and are coming close to convincing me. (But not quite... )

Anyway, I'm personally inclined to give more weight to linguistic evidence post-LoTR than to linguistic evidence of even the 1930s. For starters, it's actually quite obvious from LT I that "some rather intricate and baroque developments during the writing of GL" did in fact occur:

... the intensity with which my father used this diminutive book, emending, rejecting, adding, in layer upon layer ... the stages of a rapidly expanding linguistic conception ... GL in particular closely accompanied the actual composition of the Tales ... the languages changed even while the first 'layer' was being entered in GL ...

Now, FG being the first Tale composed means that "Rog" belongs to the first 'layer'. So already the GL entry for "Rog" must be in some measure of doubt. As Element 2 ('rog'/'raug'/Q. 'arauka') evidently came later than FG, we have further doubt that "Rog = strong, doughty" (as in FG) remains valid.

Retention of the distinct element "rog" in 1930 does not imply retention at a later date. Nor does it imply that it would not be retained. So in view of the situation, use of either argument would not be evidential. The only real "hard" evidence we have for any form of "rog", which is ultimately the only thing it can stand or fall by if we are to take a strictly authorative viewpoint, is "Balrog".

The stages of development I propose are:

1. Rog = "strong, doughty",
2. Rog = "swift, rushing",
3. Rog = "demon".

I also propose that these were successive stages, and that each supplanted the previous (with a possibility of undocumented intermediate stages).

To support this:

1. The GL definition came first, and the FG entires were the first made in it (evidential)
2. The languages developed as the tales were composed, and even as the first 'layer' in GL was composed (evidential)
3. The Tale of Tinuviel was the second composed (evidential) by which point the meaning of "Rog" had changed (I would claim this as evidential, based on (4) below, and this is what the whole thing will hinge on)
4. There is no other major word for which two concurrent but completely different meanings exist in the same language in GL or QL (evidential)
5. By 1930 we had both Rog and Balrog, but there is no evidence either way to indicate that the "rog" element was the same or different in both at this time (evidential)
6. By 1937 we had "rog" derived from the stem "RUK-", meaning "demon" (evidential)
7. There is no evidence to support (or reject) the retention of "Rog" in the Gondolin material by this time (evidential)
8. In the latter (authorative) conception, we have "Balrog" where the "rog" element is "demon", and this is the sole authorative definition of "rog" that we have (evidential)

To my mind, the argument for "Rog" being retained as the name of an Elf of Gondolin rests entirely on lack of evidence, whereas the argument for "Rog" not being retained is based on actual evidence of linguistic development in successive stages, and this makes the argument for retention weaker.