Since (at least) the publication of _Unfinished Tales_, we know that the
Telerin branch of Eldarin (which includes Sindarin and the Telerin of
Aman) shifted _kw_ to _p_. This was explicitly mentioned in the footnote
of UT:344, where we are given the example of Eldarin _alkwa_ 'swan',
which became Telerin _alpa_ and Sindarin _alf_ (transcribed _alph_), and
there are plenty of other examples and texts that confirm it.
The _kw_ > _p_ development in some Elvish languages goes back to
Tolkien's earliest works, though with variations. The language of the
Sea-elves had this feature since its very first conception, when it was
called "Solosimpilin" (cf. PE12:16-17, 21). The Gnomish/Noldorin
language kept _kw_ (spelt _cw_) in the first Grammar and Lexicon, but
it adopted the shift to _p_ during the Leeds period in 1920-5, when
Tolkien wrote the "Noldorin Word-lists" (see many of the words with _p-_
in PE13:152, and the editor's commentary on p. 134).
In the _Etymologies_, both Telerin and Noldorin (Old and Exilic) share
this characteristic. In fact we may find the same example as in UT under
the entry ÁLAK:
*_alk-wâ_ swan: Q _alqa_; T _alpa_; ON _alpha_; N _alf_ (V:348).
But there is a strange thing here: Noldorin, contrary to Sindarin in the
later conception, was not of the same linguistic branch as Telerin. On
the other hand, languages that _did_ descend from Telerin, like Ilkorin
or Doriathrin (see the trees of tongues in V:169-170, 196-197), kept the
original _kw_. In the example above we find Ilk. _alch_ and Dan. _ealc_
(where _kw_ has been transformed because of its final position). And
there are clearer examples in the entries starting in KW- (V:366):
- Ilk. _côm_ 'sickness' (< *_cwâm_) like Q. _qáme_ but unlike N.
_paw_, from KWAM-.
- Ilk. _cwess_ 'down' (the noun, not the adverb), like Q. _qesset_ but
unlike N. _pesseg_, from KWES-.
- Dor. _cwindor_ 'narrator', from _kwentrô_ (KWET-) and like Q.
_qentaro_, but unlike N. _pethron_.
And A&C, in VT45:24, provides two further examples:
- Ilk. _cwên_ 'small gull, petrel', borrowed in N. as _cuen_, but
unlike obsolete ON. _paine_, Exilic _poen_ and Tel. _pâne_, from KWÆ
- Dor. _cwend_ 'elf', like Q. _qende_ (as well as Dan. _cwenda_), but
unlike N. _penedh_, from KWEN(E)D.
Thus, I would say that the "Telerin" represented in Etym. should not be
regarded as the common ancestor of Valinórean Telerin and the
Beleriandic languages, but specifically the Telerin of Valinor, and that
the change _kw_ > _p_ was developed after they left Beleriand. The
coincidence with Noldorin in this aspect would be explained by the
approximation between Nolorin and (Valinórean) Telerin mentioned in the
_Lhammas_: "[The converse of Noldor in Valinor] was rather with the
Teleri of the neighbouring shores than with the Lindar, and the tongues
of Teleri and Noldor drew somewhat together again in those days" (V:173).
But in this case, which language developed that change originally? In
the external history of the languages, we have seen that it was first a
feature of the language of the Solosimpi (> Teleri), afterwards "copied"
in Noldorin. However, in the context of the _Lhammas_ and the
_Etymologies_ it could have happened the other way round: Noldorin "was
altered much by new words and devices of language ... invented anew by
the Noldor" (V:174).
Thus, the _kw_ > _p_ shift might have well been one of the older
inventions (other features of Noldorin, like the "Initial Variation of
Consonants" in compounds and after articles (V:298), and the plural
formation through vowel mutation, are observed only in Exilic Noldorin).
This Old Noldorin invention would have then been adopted by Telerin, as
opposed to the later development, when Noldorin was transformed into
Sindarin, a descendant of Telerin.