David, thanks for this post drawing a welcome eye to the importance of systems approaches in education!
However, I'd challenge you that just because systems are complex, a systems approach must be as well. The very exciting work in complexity theory (and the functioning of complex adaptive systems more specifically) has much to say here (the best policy-relevant summary of which I've seen is in this three-series blog post by Owen Barder: https://www.cgdev.org/blog/complexity-adaptation-and-results ). You may be familiar with Pritchett's version specific to education in his book: https://www.cgdev.org/publication/ft/rebirth-education-lant-pritchett-brief
Let me posit a further idea: What growth diagnostics did for economic growth policy (moving from Washington Consensus one-size-fits-all thinking, to a structured framework to identifying specific binding constraints to economic growth for country A in time B), we now need for education policy (and other policy environments). That is to say, we need a simple-to-use generalizable framework (a set of tests and differential diagnostics) for identifying binding constraints in a complex (educational) system.
Thoughts? Does this already exist somewhere?