I can't talk about the specifics of the software design.
But if you take into account how the Blender project is handled with consideration of how to sustain itself via the Blender Foundation, and all the related projects, you can observe how an outlier it is compared to all similar ambitious projects that are not doing so well.
This is a bad and a good sign: it makes evident how little FOSS projects take in account it's sustainability over time, on the other hand it presents an opportunity for huge improvements in this area.
It's so modular that it's very easy to extend.
A new fluid sim algorithm? A few months later Blender is shipping it.
That's also the reason it's fast. It starts in one or two seconds.