This sound change was mentioned in passing in the Outline of Phonetic Development (OP1) from the 1940s:

Also medial ĭ, ŭ usually to e, o (when unstressed) before r (PE19/60).

There is no example of this sound change in OP1, but there is one in the Outline of Phonology from the 1950s (OP2): ✶tindōmi-sel(dĕ) > tindómizel > [Noldorin] Q. Tindómerel (PE19/73). Oddly this name appeared as ᴹQ. Tindómirel in OP1 (PE19/33), though it was ᴹQ. Tindómerel in The Etymologies also from the 1930s (Ety/SEL-D, TIN). In any case, this example indicates this was a late sound change, after the Quenya rhotacism and thus possibly as late as Noldorin Tarquesta.

There are also numerous counterexamples to this sound change, such as: ᴹQ. kelure, ᴹQ. ilfirin, Q. nattira-, Q. tirítë, or ᴹQ. liruste, not to mention the plurals of words ending in i and u, such as Q. Ainur. In fact, Tindómerel seems to be the only example of the change. I see three possible explanations:

1) It could be an example of a sound change Tolkien was not fully committed to, and as such did not use outside the one example.

2) It could be very late sound change, only in Noldorin Tarquesta, and thus not reflected in spelling.

3) It could be limited in the circumstances where it occurred. Tolkien said the change applied only to “medial ĭ, ŭ”, perhaps meaning it did not occur in initial or final syllables. It may also not have applied to prefixed forms (like ilfirin or nattira-), where there would be a tendency to reform back to unmodified form (firin and tira-). This would explain all the counterexamples above except for kelure, which might just be an exception, or a reformation from ᴹQ. kelu- “(begin to) flow”.

Of the above, I prefer this third explanation.