Semantic shifts and lexification efficiency

Semantic shifts have often been discussed from a process-oriented perspective (asking what kinds of changes are likely), I will take a result-oriented perspective here, asking what kinds of meanings are likely to be lexified (i.e. expressed by roots). The functional perspective of coding efficiency leads us to expect that frequently conveyed (and predictable) meanings are expressed by short forms, and since roots must be short, we can predict that rarely conveyed meanings are not expressed by roots. For example, very few languages will have a root for 'great grandmother', while more languages lexify 'grandmother' (e.g. Portuguese avó), and all languages lexify 'mother' (except for Esperanto). This perspective seems to allow us an understanding of certain kinds of semantic shifts, e.g. that Russian mašina (originally 'machine') came to mean specifically 'car', or that German löschen came to mean 'delete (a file)' (originally it meant 'extinguish (a fire)'. Cars and digital files have become a lot more common (and frequently talked about) over the last decades, so we need short expressions for them. Shortness can come about by clipping (e.g. omnibus became bus), but it appears that it more often comes about by semantic shift. These shifts can apparently be better explained by the desired result (a short form for a frequent meaning) than by the nature of the process.