How do ancient vowel suppression and prothesis work?

There are several cases where Sindarin words seem to have lost internal vowels in a way unrelated to the loss of morpheme-boundary vowels in Middle Sindarin, and more reminding of the syncope processes seen in Quenya and Doriathren. For instance

PQ *Tamana > CS Tavn

ANAD > CS And

PQ *Talata- > OS Taltha-

(PQ *Oronē >) MS †Oron > CS Orn

to name just a few. At first I thought Sindarin was influenced by Doriathren/Quenya and randomly employed the syncope for some words while not doing so for most others; however, this doesn't hold up for several reasons. First, there are even (late) Primitive Quendian (or maybe early Common Eldarin) terms exhibiting a similar vowel loss, like

CE Ostō < PQ *Osotō < SOT,

CE As'tāră < PQ *Asatāra < SATAR,

CE Alkwā < PQ *Alakwā < ALAK,

CE Stalgondō < PQ *Stalagondō < STALAG

and it's even represented in some roots:

MIZDI < MISID

ANDA < ANAD

and unlike the usual syncopes of Arda, not all lost vowels have a similar vowel in the preceding syllable.

CE Nki < *Niki

CE ū̆k’lā < PQ *ūkalā < KAL

CE K'lā < PQ *Kalā < KAL

So, this seems to be not a syncope as in Doriathren/Quenya/etc. and it's also not random. For instance, roots of the form (C)VnVd always lose their second vowel in the extended form, e.g.

CE Kwendī < KWEN(ED)

CE Wendē < WEN(ED)

CE Spindē < SPIN(ID)

CE Andā < ANAD

Now, if there is a set of consistent rules behind all this, I can't see it, and maybe I'm just messing three different processes here, but can anyone tell me what's going on with word-internal lost vowels? How can PQ *Tamana become CS Tavn, while PQ Lamana becomes CS Lavan?

PS. I couldn't find an Eldamo article on vowel-suppression in the Primitive Elvish Phonetics section, but I think there ought to be one. The process can't have been in the language from the very beginning, because the vowel-suppression seems to have taken place after certain ancient changes like:

V̄CC > VCC (else ū̆k’lā would be uk'lā),

VyC > ViC (else PQ Yuyuñal would be Q Yúial),

VŋV > VɣV (else PQ Yuyuñal might never have lost ñ, since y is a C and it's not said that CŋV > CɣV)

Telerin cl > cul (else PQ *Acalar, yielding S Aglar, would become Telerin Acular instead of Aclar (cf. Hecul))

On the other hand it must have been a fairly ancient process, occuring before other CE developments like:

sd > zd (CE Ezdē might otherwise be Esdē. Also MISID > MIZDI and not MISDI)

CE loss of -a/-e/-o (cf. As'tāră, and CE And(a))

CE change of -i/-u to -e/-o in polysyllables (else Nki would be Nke)

PPS. By the way, why is it necessary to add a prosthetic to KALAR in order to deduce a word like Akalara from it? Wouldn't Kalara do as well? Since when do we need more than a root and a suffix to form a word in PQ? And is this prosthetic maybe responsible for the lost vowel in Ak'lara, or is the occurence of both a coincidence?

Thanks in advance for your answers.

That is an exceptionally complicated question. I agree that Eldamo should have descriptions of more ancient phonetic changes and word formation rules, but I haven't gotten around to writing them up yet.

You are having difficulty finding a coherent set of rules because there aren't any. The change of (for example) ANAD > andā isn't a phonetic development. It is following one of various possibly word formations from the ancient root.

Roots like KALAR are not words. They are expressions of set of characteristics for a collection of related words, consisting of a base vowel (A) and one to there consonants (K/L/R). From these are various possible ancient words that may be formed from these roots, like kalar, aklar, aklāra (> Q/S. calar, Q/S. alcar/aglar, S. aglor). There is no fixed pattern because different roots may have a variety of different words formed from them.

Tolkien talked about the various word formation rules in a document labeled Tengwesta Qenderinwa (two versions, one from the 1930s and another written around 1950): PE18/85-98. For example, aklar is formed KALAR using a vocalic augment with suppression of the first interior vowel (a-k’lar). These are word formation rules, not phonetic changes.

Hopefully that’s enough to give you a general idea. See PE18 for more details.