I brought up Similar Place Avoidance (SPA) a couple of posts ago. Here is a neat case study of it in action, one that I have already noted quite some time ago.
The Finnic languages are usually considered to have no strict noun/adjective division, and adjectives are analyzed as the same part of speech as nouns. But this does not mean that there would be no visible differences between words that have regular nominal semantics (“substantives”, in the Finnish grammatical tradition) and those that have adjectival semantics. While there are a couple of underived, bare-root adjectives (e.g. kova ‘hard’; nuori ‘young’), most Finnic adjectives are marked with an ending that reveals their semantic function.
In Finnish one of the more common adjectival endings in “adjectival roots” is the suffix -ea, -eä (in the forthgoing marked together as -eA). This is in contrast to suffixes that derive adjectives from words that, as their bare root, function as nouns. E.g. punainen ‘red’, with the highly common adjectival ending -inen, derives from a separate noun puna ‘redness’, whereas valkea ‘white’ does not allow a synchronic morphological division into a self-standing root with one meaning + a suffix with another. 
There is however a curious statistical gap in the distribution of -eA: it seems to shun preceding dental consonants. Using the Nykysuomen sanalista wordlist as a reference, there are no Finnish adjectives ending in -neA to be found, and no more than six ending in -teA, either:
- kiinteä ‘solid’
- kostea ‘moist’
- lattea, litteä ‘flat’
- nuortea ‘youthful’
- pirteä ‘cheery’
- reteä ‘chill’
A seventh could be implied in vetreä ‘spry’, which seems to come via metathesis (perhaps by influence of potra ‘thriving, brisk; usually only in potra poika‘?) from earlier *verteä. Compare verrytellä ‘to stretch, to flex’, which appears to share the same stem /vert-/. 
For comparison, we can count e.g. adjectives ending in /-peA/. There are more than twenty of these:
- apea ‘sad’
- hempeä ‘romantic’
- hilpeä ‘jocular’
- hulppea ‘extravagant’
- kalpea ‘pale’
- kapea ‘thin’
- kepeä ‘light’
- kipeä ‘sore’
- kirpeä ‘sour, crisp’
- kopea ‘arrogant’
- leppeä ‘mild (of weather)’
- nopea ‘fast’
- nyrpeä ‘grumpy’
- rapea ‘crisp, crunchy’
- ripeä ‘prompt’
- suopea ‘benevolent’
- suppea ‘concise’
- turpea ‘swollen’
- tympeä ‘stale’
- upea ‘fantastic’
- ylpeä ‘proud’
/t/ is, overall, a more common consonant than /p/ in Finnish, so getting this kind of a result is not a priori expected.
I have also counted examples with other consonants. An uneven distribution clearly biased against dentals continues: there are e.g. on the order of 60 adjectives ending in -keA, and about 20 in -meA.
The result is however quite understandable in light of the SPA principle. The Finnish adjective ending -eA comes from earlier *-eðA < Proto-Finnic *-edA < Proto-Uralic *-ətA. This would be the easiest to demonstrate using the peripheral Finnic languages Veps and Livonian, which retain PF *d, but even some examples older yet can be found:
- Fi. dialectal kalkea ‘hard’ ~ Moksha /kalgəda/ ‘id.’ < *këlkəta
- Fi. tankea ‘stiff’ ~ Moksha /taŋgəda/ ‘id.’ < *tëŋkəta
- Fi. oikea ‘right’ ~ Moksha /viďä/ ‘id.’ < #wɜjkəta
- Fi. pimeä ‘dark’ ~ Komi /pemɨd/, Udmurt /peĺmɨt/ ‘id.’ ~ Proto-Samoyedic *pəjmətä ‘id.’ < #pid₂mətä
So we can expect SPA to have intervened, over the course of millennia, to somehow clean out any undesirable sequences of two syllables beginning with dental stops. Of course, the modern Finnish ending has no signs of a dental element anymore, and so we could perhaps hypothetize that the six or seven exceptions have been formed only after the loss of the segment. (Indeed, as far as I can tell, none of them have exact equivalents in any related language, and most are a somewhat limited even in their distribution across the Finnish dialects.)
But how has this worked exactly in practice? There is no shortage of Finnic word roots with medial -t-, and a fair number with medial -n- as well. Does this imply that words that once upon a time ended in *-tedA and *-nedA have changed to something else?
I believe the solution has been instead morphological. As an adjectival ending, -eA still has several competitors in Finnish, and every so often we can find sets of essentially synonymous adjectives (with only minor differences in register and tone) that differ only in what suffixes are employed. Examples that can be noted in modern Finnish include -Ut (as in kevyt ‘light’), -(A)kkA (as in kalvakka ‘pale’, rivakka ‘prompt’), and participles such as the past active -nUt (as in turvota ‘to swell’ → turvonnut ‘swollen’).
One suffix that comes particularly close to -eA in shape is the regular present active participle -(e)vA, also commonly repurposed for deriving adjectives.  OK, the preceding vowel is taken from the verb stem and is not a part of the suffix: mene- ‘to go’ → menevä ‘going; busy’, but osu- ‘to hit’ → osuva ‘hitting; apt’, or paina- ‘to press, to weigh’ → painava ‘pressing; heavy’. But the interesting part is the existence of a couple of adjectives that seemingly possess this ending, and yet are not derived from any known verb. Frequently they seem to derive from nominal stems instead. What is more, quite a few of these are both apparent e-stems, and have a preceding /t/:
- etevä ‘skilled’ ← esi : ete- ‘fore-‘ (but not ‘to advance’)
- harteva ‘wide-shouldered’ ~ hartia ‘shoulder’ (no verb ˣharte- ‘to be shouldered’ exists)
- jäntevä ‘wiry, spry’ ~ jänne ‘sinew’ (no verb ˣjänte- ‘to be sinewy’ exists)
- kalteva ‘slanted’ ← kalte- ‘side’ (but not ‘to slant’)
- kätevä ‘handy, dexterous’ ← käsi : käte- ‘hand’ (but not ‘to do with hands’)
- lehtevä ‘leafy’ ← lehti : lehte- ‘leaf’ (but not ‘to be leafy’)
- luonteva ‘natural, easygoing’ ~ luonto ‘nature’ (no verb ˣluonte- ‘to be natural at’ exists)
- ponteva ‘vigorous’ ← ponsi : ponte- ‘motion, exertion’ (but not ‘to exert oneself’)
- roteva ‘robust’ (seemingly underived)
- varteva ‘tall (of people)’ ← varsi : varte- ‘stem’ (but not ‘to be tall-bodied’)
I also have counted a couple of cases where this kind of suffixation seems to have taken place before non-dental consonants, but these are clearly rarer. There is only one debatable case with -pevA: lipevä ‘slick, unctuous’. The bare root lipe-, used as a base for a large number of words related to slipperyness, is not verbal, no, although there is the quite close-by verb lipeä- ‘to slip’ (its regular present active participle is lipeävä). With -kevA there are six cases (e.g. väkevä ‘strong’ ← väki : väke- ‘people’ < *’power’). So in the end, probably the extension of -(e)vA from a regular participle function to another adjectival ending has taken place here as well. But we can still see a clear discrepancy between the 10 : 7 ratio of -evA to -eA adjectives when the previous consonant is /t/; a 6 : 60 ratio when it is /k/; and a 1 : 21 ratio when it is /p/.
What have we seen, and are able to conclude, so far?
- The Finnish adjectival ending -eA has been disproportionally rarely applied to stems that have a medial dental stop.
- By contrast, the ending -(e)vA has been disproportionally often applied to stems that have a medial dental stop; and, arguably, disproportionally rarely to stems that have a medial labial stop.
- These results support viewing Similar Place Avoidance as a potential statistical linguistic universal.
- The ending -evA has probably been originally extracted from the participles of e-stem verbs.
- This extraction may even have happened specifically to acquire an alternative for -eA.
…and skipping a bit further ahead of syllogistic step-by-step argumentation: the most general statement of what is going on here is that derivational morphology is not random. In a morphology-rich language, affix alternants and synonyms will form an “ecology” where potential words are selected for according to their adherence to some kind of aesthetics, such as phonetics-rooted criteria. SPA is one example of such a criterion. There are probably several other relatively general ones that could be identified crosslinguistically. And indeed, there are some other examples that I could illustrate as well.
Some further questions
As for this particular case study: so far I’ve only shown that one particular Finnish adjectival suffix has a non-random limitation on its occurrence; and identified only one other suffix that has been taking on the work of -eA. There would likely be others as well. Most of my -tevA examples were, in the end, derived words, based specifically on e-stem nouns. So what about primary adjectives? Or adjectives derived from A-stem or O-stem nouns? Do they perhaps also have their own specific preferred adjectival endings? I don’t quite have an answer yet.
Also, what about the other Uralic languages? How have they solved this issue? A couple of the adjectival stems on showcase here have cognates elsewhere in Finnic, too (e.g. kätevä being also found in Karelian). But since the adjectival suffix *-ətA dates already to Proto-Uralic, we can expect this particular problem to come up several times before as well. Could we find similar limitations in its distribution in e.g. the Mordvinic or Permic languages?
Obviously this question could also be extended to any other suffix type. Deverbal nouns? Frequentative verbs? Deminutives? There’s a lot that could be studied about derivational morphology across the Uralic languages.
 Historically, there may well exist a derivational relationship, though. Note e.g. valo ‘light (n.)’, vaalea ‘light (a.)’.
 The ultimate root for these could be veri ‘blood’, the implied derivation being thru an unattested (?) verb ˣver-tä- ‘to be full of blood’ > ‘to be energetic’?
 This even has develop’d historically by a similar intervocalic lenition from *-βA < *-bA < *-pA, so at almost any given stage of Finnish prehistory stage it would have been the exact [+labial] counterpart of the [+coronal] *-ətA.