My last project - For SCIENCE! - developed in a pretty straightforward manner. I put it on pause mostly because it turned out to not really be any fun. I learned a lot from it, and starting up my next project really punched me in the face with those lessons. So let's talk about it again!
The reason For SCIENCE! was so easy to develop was because my mood was strong throughout. I never felt uninspired by my game. And a big part of that was the content. By buying content from the asset store, I filled my game with detailed, fully-textured models right from the start. Even in its early stages, every test run felt like I was building something. A lot of devs can get away with placeholder art, but I evidently can't.
When I moved on to Gravity Grain, I knew I'd need a lot of content, so the first thing I did was create a content creation system. But this really distracted me from creating the game. As I should have realized, content creation tools are really something separate from the game. In fact, the concepts are so distinct that I literally packaged my voxel-object tools up as their own library, so they could be used and reused and imported and exported into any game, anywhere.
... Why not just import actual models, at that point? Why not just let people make great stuff in Blender or Maya and import it through the standard Unity asset pipeline? Why create a complicated, difficult-to-maintain voxel system?
Sure, the voxel system had some theoretical advantages. Shared materials between all in-game objects. Easy and meaningful deformation and damage. But... it's a TON of work, and the result has a very specific kind of awkwardness even with the cool smoothing mechanics. It would be more rewarding in the long run to give the players and developer(s) the kind of freedom you get with professional modeling tools.
So, I'm throwing it away.
I'm going back to the proven method I used in For SCIENCE!: getting third party content and feeling like I'm making progress rather than struggling to create both a tool and a game simultaneously.