The only reason I see is when we look at what are the apps added per server-group, it seems to be picking up the information from server data directory.
The deploy command under server groups or the equivalent "enable" on web-consile seems to copy this content (from server data directory) to the server/tmp/vfs directory and extracting and using it from there.
Like you say it is duplication of data in multiple places.
Actually I dont think the data directory is part of the config.
The data directory with the hashed files is not scanned. If a file is there and the domain.xml does not point to the hash, it is ignored. And multiple profiles/groups can point to different combinations of artifacts. I can understand to have one replicated content per host (directory), but not per server. (I havent checked AS8 if this is still the case, but if yes I guess I will open a Jira).
This sounds like a reasonable enhancment,
as there could be some performance and disk space gain when there are big deployments or many of them...
so please file jira for this.