I support the idea of building from git master, but that reason for that is (as demm described) how KDE devs work. As I see it, it steals unnecessary time from demm: rebuild on new Frameworks, additional patches (back)ported to stable from master git to remove problems, ... With building from master demm has a less work to do (and more time to focus on other nicer things 🙂). If the tarballs were more tested, then it would be another thing.
Plasma 5.8 will be an LTS release (the current master in git, release in mid October). Is there a chance that at this point the devs will run 5.8 or will they stay on master (future 5.9)? I don't know the KDE development good enough to make a guess on this. If the devs switch to the LTS to work on and 5.9 becomes a lame duck, than the whole point of switching to master might backfire. There is currently no date set in the release shedule for Plasma 5.9 (which should be released around jan/Feb 2017).
Also, what do you want to build from master git. Plasma (Desktop), Frameworks and Applications? Or only Plasma and Frameworks and Applications stay on tarballs?