It's a nice structure but based on my experience I wonder how you address these conditions:
A. Your installs all appear to land in the same destination. If a user needs to install all 3 versions at once will your process allow this. Can they unambiguously tell which version of every executable they are triggering?
B. How do you handle new files that exist in TEST and/or TRAINING but not yet in LIVE?
