Content Management Revamp

For me, currently the weakest part of Blender is scene assembly. It fails miserably when it comes to managing large data sets, partly due to the lack of features, partly due to the lack of logic. Here are my proposals:

1. Outliner hierarchy should reflect parenting - so by drag and dropping you change the parent-child relationships. Properties of father can be inherited by children (modifiers/materials/transforms ) - so the overrides system comes into play here.
2. Collections are basically still groups but nicely packaged into outliner. They are heavily underutilized. Basically these should be just nulls and their transforms should matter (default at 0,0,0). Why make life harder? Take Unity for example.
3. Instancing is a mess and works 1 level down, so instancing instances fails. The most frustrating part here is that the user needs to waste time on checking which config will work in many situations. Why instancing via PS loses PSes on instaned items? Why Particle Instancer works only for single objects? If we treat collections as null, then I guess we can simplify things here.
4. manual creation of proxies for each member of a collection - well that's a big drawback.
5. proper large-scene assembly tools would be a great addition. So point cloud generators, objects scatterers taking into account AO, other objects proximity, implicit colliders, defined area types (simple color coding), grease pencil, curves, populators based on PCs, vertex colors, etc - working in a logical way, instancing down the line to the lowest level of hierarchy. Not some makeshift particle system workarounds. Take Clarisse for example.

These are only some ideas which I am sure would require a great overhaul of the code, but I guess much more important than massaging UI for the nth time. I really do appreciate 2.8 effort (it should be called like 3.0), but this is Blender's time - we need to make it PRODUCTION READY in order to dethrone the old fat so-called "big" 3d software packages.