All this talk of header IDs and ETE reminds me of my experiment last year encoding the structure of things in IDs. The idea was to use deepening numbers similarly to how many product databases do it – 1000, 1100, 1103 being related products. Here’s my conclusion to that thread:
Mostly induced by @johnnydecimal coaching me back from adding this complexity 
I wanted to try this out partly because of seeing header IDs. So I find it interesting that they are still causing restlessness.
@ontologist has brought some excellent professional wisdom to bear here. With my experiment in mind, I would emphasize: JD is about making things findeable, and it works because the categories you create are fake. As soon as you start encoding real-world structure, your filing system’s days are numbered. I suspect that will be confirmed experimentally as a Law of Information Management at some point if it hasn’t been already.
I agree about the tags or groups. As long as it’s a closed set, itself crafted in an intentional way.