Good Linguistics

Linguists have good reason to be very cautious about putative etymologies: Some words exhibit such a complicated history that any '(un)educated guesses' or phonetic alignment would surely lead to completely wrong conclusions: Take the word lord<lourde<lowerd<louerd,loverd,laford<lhoaverd<hlāford < hlāfweard : hlāf “bread” + weard “guardian” : loaf-warden if you will. Without historic background any etymology would certainly be misguided, the correct derivation impossible to guess.

azerbaijan Āturpātākān Atropates,[32][33] a Persian[34][35][36] satrap

There is a gradient between words which possibly should be connected on wiktionary, but are left out because of doubt and words which should be connected but are actively rejected because of dubious arguments and "formal difficulties".

Sometimes Wiktionary surprises with Good Linguistics, opposite of Bad Linguistics

Good Historians:

«The field is becoming much more open to the idea of using terms like "probably, likely, possibly"»

Good Linguists acknowledge, that similar to genetic admixture, creole languages inherit traits from both parent languages. Better Linguists acknowledge, that in fact most languages should be considered Creole languages forming a continuum between neighbours.

https://en.wikipedia.org/wiki/Mixed_language

One of the axioms of revised language theory is:

If words from the same language familie(s) look similar on the surface they are most likely connected, even if the reconstructions seem to be divergent on first examination.

Examples at missing connections

shell shelter shield https://en.wiktionary.org/wiki/shoulder#English
but not shell scallop skull σκληρός "hard" σκελετός σκέλλω (skéllō, “dry“)
causative of halt ⇨ s'halt : shelter / shield

gar(ready) gären" Garum, Vinegar ⇔ a'gjar sour sauer @ jar < Arabic جَرَّة (jarra) < jaraf,ḡurf,√saraf "drink/liquid…" super-root > karaffe carafe
غُرْفَة (ḡurfa, “cup or dipper”, from غَرَفَ (ḡarafa, “to ladle”
slurp Schlürfen ⇔ aQua.phore
S-*-r-f :
𓋴 𓅨 𓂋 𓇋 𓈗 𓀁||drink
𓋴 𓏴 𓂋 𓇋 𓈗 𓏲 𓂧 𓆑 𓂻||drink

ش ر ف sharif honor, noble @shar => sir ≠ sheriff (shire + reeve < scīr ġerēfa:(Kur)Graf 𐌲𐌰𐌲𐍂𐌴𐍆𐍄𐍃 γρᾰ́φω )
sounds reasonable. If not related, the term was likely enforced by semitic/hindi... speakers

Downward sound change

In by far the most cases sound-change happens from high entropy to low entropy.
osteo < 𒄩𒀀𒀸𒊭 /ḫāssa/ bone *h₃ésth₁ < Кость < 𓈎 𓋴 𓌟 qʳstwo
If an upstream word appears in the tree of related cognates, it has to be taken as a root form!
Especially if it is explicitly attested in the historic record (as is the case with Кость < 𓈎 𓋴 𓌟 qʳstwo )

enforcement

if a native community has no preference for competing words than foreign communities influence and select the preferred word which has the closest perceived relative in their native language. sheriff (shire + reeve) ⇔ sharif Might be an example

avoiding directionality

Good-Linguists] treat these forms equivalently: crypt κρυπτή (kruptḗ), female form of the adjective κρυπτός (kruptós)

In the lack of contemporary old Proto-Germanic documents it's near impossible to say whether a term entered Germanic from Latin or the other way round.

onomatopoeia

The question of onomatopoeia is near orthogonal to the question of relatedness: There are hundreds of plausibile sounds connected to entities. If the whole planet uses different sounds to describe an object and (as an example) two neigboring languages use “pop” there is a high chance that these two “pop” are related, onomatopoetic or not.

other good linguistics:

A very useful contribution is the collection of corpora:
corpus of linear A https://sigla.phis.me/#doc:HT%2012
demotic

Agade good kyau.de kyau.ma mai kyau make eu

Wave theory and feature distribution

https://wals.info/feature e.g.
https://wals.info/feature/7A#2/4.9/149.8

The balance between false (folk) etymologies on the one hand and false rejections on the other hand is extremely delicate. On elucidation of this eternal problem could be cluster-analysis and unbiased alignment. Even though some connections may forever be unprovable, it doesn't harm to put two similar forms next to each other, if there is a probable historic cause to them being related or at least influenced after all.

Point in case:
Linguistics spent decades extracting seemingly different roots for English flow and Latin fluvio, Latin Theus and Greek Deus/Zeus. It's time for a new generation of Linguists to reconnect these different roots further down in the tree.