I think the blender secret sauce is their artistic projects.

Put a bunch of artists in the same room as the developers and have them produce a work.

It ferments this amazing combination of aggressive QA testing(the artists) and top tier technical support(the developers) while focusing on real problems(the work) that really brings out the best possible product. The GIMP project would probably be better off if they invested in a couple of rounds of this.

I think blender always had an amazing, ahead of it's time interface. It did lean overly hard on knowing the hot keys, probably a product of it being an in house tool but the opensource versions have worn a lot of those rough edges off(menus to provide clues) while keeping that same super smooth workflow core.

>It ferments this amazing combination of aggressive QA testing(the artists) and top tier technical support(the developers) while focusing on real problems(the work) that really brings out the best possible product.

Devs stil run the show, and it can be a huge effort to convince them to change course on something. People were complaining about the mouse controls for years and it took until 2.8 to finally convince the org otherwise to adapt a more typical workflow.

It's a rough balance because the other extreme is artists making unrealistic demands based on how the system is architected (or worse, the deceptively simole: https://xkcd.com/1425/). So I don't really have a solution.