Seems to me there are two tracks by which words get imported:
"paamara" and "pandita" (no value judgement intended). The paamara
imports are fully pragmatic, and sometimes follow improbable
associations. To take a very artificial example, suppose the arts college
building of the university has a well-known beehive fronting it; an
illiterate rickshaw-driver might identify it as "tayne-paTTU kaaleji",
and by association, liberal arts studies might get translated over
time to "tayne-paTTU chaduvulu." Other, more intuitive transformations
might involve well-entrenched mechanisms, such as a physical
description of the object's shape and behavior. Interesting examples
for this come from the world of fireworks: "Tapa-kaaya"=onomaetopic
"Tapa" sound + the common word "kaaya" for any sealed container. My
favorite is the word for rocket: "taara-zuvva=taara+chuvva=a thin,
short rod-shaped object that reaches for the stars (Or, does the
"zuvva" refer to the sound it makes?). In Telangana, "sura-sura-batti"
is used for the common sparkler, indicating the sound it makes, along
with its shape and function (gradual burning). On the coast, it is
"kaakara-puvvu-vatti", a variant of the same concept.
Another "-kaaya" example I like is taabeTi-kaaya for water canteen,
which is a perfect description of the object in question. As far as I
know, these highly sophisticated constructions are purely non-literate
and spontaneous; no writer or committee thought them up.
Another example is the turn-of-the-century use of "batikina kaaleji"
for the Madras Zoo and "chacchina kaaleji" for the Madras Museum.
In some sense, the above technique is "semantic" in nature; there is
some preservation/transformation of the meaning of the concept into
the native idiom.
Another "paamara" import channel I can think of is the literal or
syntactic adoption of the foreign word, but subjecting it to the
"naluguDu" process in native tongues. Some years back, PALANA-garu
posted some examples of SrikakuLam dialect that was rich with such
imports, from English as well as other languages. I suspect that the
use of "seku-hara" for sexual harassment by the Japanese falls under a
conscious use of this track of adoption.
Without wishing to be romantic about this, I think the common feature
of "paamara" imports is what Suresh-garu noted: they harmonize well
with the language, and will work well for poetry. Professional
linguists will probably tell us that such "paamara" adoptions are an
important aspect of creole formation, which in turn is a key aspect of
language development.
The "panDita" track is more deliberate in nature, and has the
disadvantage that it is confined to the literate elite--disadvantage
because of the relatively small number of persons involved in the
import, which means the statistical smoothing effect of "naluguDu"
(syntactic as well as semantic) is that much less powerful. Also, the
prevalence of centralized mass communication techniques such as
printing, cinema, broadcasting etc. means that there isn't as much
time or opportunity for cacophonous constructions to get weeded out
through natural selection, as would happen in the "paamara" track.
What this means is that, when it comes to translations and imports,
the language is hostage to the talent and "ear" of the person doing
the importing.
I think it will take a long time to fully understand the impact of
modern "panDita"-track imports on the language, since mass literacy
and mass communication are relatively new phenomena to Telugu. Maybe
with the wider spread of literacy, education and mass communication
tools, the "ivory tower" effect of today's "panDita" transformations
could be neutralized to some extent, moving back to a literate version
of the more natural "paamara" transformations. Maybe even an entity
such as the internet can play a role in the "democratization" of the
evolution of the sishTa-bhaasha.
Bapa Rao