Finnish armas ‘dear’ has a somewhat interesting etymology: the word is considered to derive by borrowing followed by semantic amelioration from Germanic *armaz ‘pitiful’.
If we were given no other data, this argument would have to remain rather hypothetical. The shape of the word does suggest an Indo-European loan, but allowing major semantic drift as a free assumption is an easy excuse for finding Germanic or Baltic etc. loan etymologies for almost everything in the Finnish lexicon (and given some effort, I’m sure the method could be also stretched to prove that Finnish is actually, say, a highly divergent dialect of Chinese). Aluckily, according to the data reported in SSA , there are several additional pieces of evidence that point to a meaning ‘pitiful’ having existed in Proto-Finnic as well.
- A “dialectal” meaning ‘pity’ is attested for the Estonian cognate (equivalent? ) armas.
- The derivative *armas-ta- (> Fi. armastaa) has pity-related meanings somewhat more widely, in Karelian, Estonian and Livonian.
- Finally, an exclusively pity-related parallel derivative from the same root appears to exist: armahtaa ‘to pardon, have mercy on’ — though attested slightly less widely: this appears in most northern Finnic varieties other than Veps, but in southern Finnic in only Votic, and it’s possible though not strictly required to consider it an Ingrian loanword in there.
The last one of these I actually find more interesting yet, though for a different reason. Namely, why does -h- appear here? It is true that the stem of armas in inflected forms is *armaha-, as in the genitive singular *armahan > Fi. armaan (~ armhan in Veps or Kven); but a stem vowel *a is not normally lost before the verbalizer *-ta-, and no consonant stem **armah- exists for this declension class.
One explanation might be haplology. Lauri Hakulinen seems to implicitly suggest this solution in SKRK,  listing the word under verbs derived by the momentane suffix *-ahta-. I.e. ‘to suddendly pity’ = ‘to pardon’? After this we’d have to assume contraction of the somewhat awkward stem *armahahta- to the attested armahta-.
This approach however suffers from the problem that Finnish momentane verbs are productively derived only from verb stems, not from nominal stems. Hakulinen only reports five other verbs formed in this fashion. Two of them actully derive from original *-eh-stems and they might be simple *-ta-derivatives after all (repalehtaa, roikalehtaa ), and for other two, derivation from a verbal stem does not seem to be possible to rule out (riemahtaa ‘to rejoice suddendly, erupt in celebration’, tipahtaa ‘to drop suddendly’ ). This leaves vapahtaa ‘to redeem, liberate’ (← vapaa ‘free’) as the only clear parallel.
Now vapahtaa is of course semantically very close to armahtaa, and this seems like a good reason to suspect that they may have affected one another’s formation in some way. Comparative examination however suggests that it’s probably armahtaa that is the model, and vapahtaa the remodelled verb. As mentioned, the former has cognates in multiple Finnic varieties; meanwhile the latter is restricted to Finnish. The root *armas also seems to be a relatively old Germanic loan, being found everywhere across Finnic, while vapaa < *vapada is a more recent Slavic loan, absent from marginal varieties such as Veps and Ludian.  So we have no clear solution here for armahtaa.
I have a different hypothesis in the works, though, that seems to fit in here quite well.
An interesting gap of general Finnic morphophonology is that no words ending in *-ah can be reconstructed for Proto-Finnic, and to my knowledge no corresponding declension can be observed in the modern languages either. This contrasts with a large number of words of the *armas type, ending in *-as : *-aha-; and an equally large amount ending in *-eh : *-ehe- (directly attestable in Karelian: hameh ‘dress’, veneh ‘boat’, etc.) A couple examples of *-es : *-ehe- exist as well (Karel. kirves : kirvehe- ‘axe’), and one or two cases of *-oh (Karel. orih ‘stallion’). Frequently we can also find among these “sibilant-final” words  discrepancies between the Finnic languages in the stem type: e.g. Finnish helmi ‘pearl’, a bare *-e-stem, corresponds to an *-es-stem helmes in Estonian, and an *-äs-stem ēļmaz in Livonian.
This makes me suspect that at some point in Finnic prehistory, general morphological levelling may have taken place here; that at one point, stems with a nominative *-ah existed as well, but these were later all reassigned as either *-as-stems or as *-eh-stems.
This is structural speculation so far. But I think there is at least one good reason to suspect the former existence of a class of *-ah-stems: in old enough Germanic loanwords, *s/*z are quite regularly substituted by pre-Finnic *š > Late Proto-Finnic *h. In the case of *-eh-stems, we can indeed find some direct correspondences of this stem type with the Indo-European masculine nominative singular ending *-s (> Germanic *-z): e.g. the above-mentioned *hameh ‘dress’ from PGmc *hamaz, or *padeh ‘path’ (> Karel. pajeh : patehe-) from PGmc *paθaz.  But *-as-stems arrive on the scene seemingly quite early, sometimes even in parallel with an *š-substitution: Fi. keihäs ‘spear’, hidas ‘slow’, from PGmc *gaizaz, *sīθaz! This all would surely be easier understandable, if we assumed for early Finnic declension patterns such as *keišäš : *keišäšä-, *šitaš : *šitaša- > ? *keihäh : *keihähä-, *hidah : *hitaha-, later levelled to the directly reconstructible *keihäs : *keihähä-, *hidas : *hitaha-.
I do not yet have a clear enough grasp of the overall picture though to say if the levelling process might have been regular in its output anywhere in the Finnic area — or even, if this should be assumed to have been a pre-Proto-Finnic or a post-Proto-Finnic process. 
But armahtaa seems to regardless fit into the framework quite nicely: the word would turn out to be after all a simple *-ta-causative, only one based on a now-lost consonant (nominative) stem *armah! The semantics also fit this picture: as noted above, armahtaa is an exclusively pity-related verb, with no associations of love. Noting again the semantic trajectory of the basic root word — Germanic ‘pitiful’ → presumable earlier Proto-Finnic ‘pitiful; dear’ > later Proto-Finnic ‘dear; pitiful’ > modern Finnic ‘dear’ — this verb was thus probably formed at an earlier time than armastaa, perhaps before the development of the meaning ‘dear’ entirely.
 “SSA” and “SKRK”, two indispensible sources in the study of Finnish etymology and morphology, have now been added to my Bibliography page.
 I sometimes feel that a pair of words in closely related languages that have the exact same meaning and shape should perhaps be described in stronger terms than “being cognate”. Given that we are usually comfortable saying that a given word “exists”, as a single entity, in several distinct dialects — and that the language/dialect distinction is arbitrary — it might be useful in an etymological context to claim that e.g. English mouse and German Maus are not merely “related”, but in fact the exact same word, just spelled in two different ways. This issue comes up the most often in etymological dictionaries, where a traditional “every language has distinct words” approach will sometimes lead to heavy repetition: “Finnish armas is cognate to Ingrian armas, Karelian armas, Estonian armas, Votic armas…”
 Though neither of these is familiar enough to me that I could do a closer semantic assessment of this solution.
 These seem like they would likely be derived from riemuita ‘to rejoice’, tippua ‘to drop’ rather than the bare roots riemu ‘joy’, tippa ‘drop’.
 Of course, currently Veps and Ludian are anything but marginal when it comes to Slavic contacts; but the oldest Slavic loans in Finnic appear to predate late Proto-Slavic proper (this one as well: reflecting early PSl *svabadā rather than late PSl *svoboda), and they were probably adopted from the archaic Old Novgorod dialect, with a main contact area close to Ingria and the Pskov region.
 Recall that Finnic *h < *š.
 I do not recall offhand what is the standard explanation of the 2nd-syllable *e of these, though.
 This might even have a few repercussions for Finnic historical phonology, but I will refrain from going into the topic for now.