Yes, that why I imagine a “neutral” model. I don’t favour any approach. I want to add Notes and the real job will be in the notes. This can already be done for Places: there is a “central” record (the place) and you can add notes to explain the name (and alternate names). I’d like to have the same possibility, which requires a “central” record for an abstract name (grouping all variants)
As an example of this ambiguity, I’ll take Cohen and Berger.
Cohen can be either an Hebraic name (a priest) or a Breton name (oak-tree). Therefore you need two Group as keys Cohen (H) and Cohen (br).
Similarly, Berger may come from Latin (shepherd) or from German (mountaineer), requiring again two Group as keys Berger (fr) and Berger (de).
The group record can be used for any purpose: linguistics as shown, geographic to exhibit independent origins, …
This is another aspect which is not basically related to Group as since it can be used (and is used) even without Group as. And this is a very cultural-sensitive point (see the recent question about Ethiopian names.
I am only in the preliminary steps so any suggestion is welcome.