OSM decided to play on safe size with regards to potential copyright issues.

CC-BY license certainly applies when a substantial amount of data from Geonames ends up in OSM.
If you take over a place name in a romanized form it is quite likely to be able to detect that the source of it was geonames, based on the creative way romanized ways are often created.

You are right that factual data can’t be copyrighted. A collection of facts (legally called a database) can be protected and is protected by database rights for example in Europe.

The preferred way of getting data into OSM is on-the-ground survey by a local community. That is a major benefit of OSM compared to other data providers. Only the local community can keep the data updated.

Even with it being tempting to improve OSM faster by doing imports, it has certain drawbacks we need to be aware of.
I might point to the US tiger import. This is considered to be one of the major reasons why there is no such strong OSM community than in Europe.

Have you ever spend an evening correcting spelling mistakes in Thai script? Not as rewarding as mapping a new place out of nowhere.
As of today still over 100 ways with a wrong spelling of “Soi” around:
http://reports.osm-tools.org/index.php?report=6

Proper fixing would be to add name of the street as well, not just fix “Soi”. Only search/replace is not enough. I could have done that easily myself.

Stephan