Sure! I am coming from sphinx, another python based documentation tool that is used for example for the Linux Kernel docs, Python's docs, etc.

While it is great for documentation, we used it for the whole website of a project, mainly because people in the team already understood it. But we ran into many issues when it came to adding a blog...

Sphinx has a hard requirement for input files to be on disk. It means that, for example, it would be hard to add a page that lists all blog posts posted in a certain category (as the "meta" page would have to inspect other pages and then be generated on the fly). The only option is to pre-generate such pages into the source folder before a build.

I think that the inputs should be abstracted in a way such that there isn't a hard requirement on the filesystem. For example, extensions should be able to add input files using code, without having to write them to the filesystem first. This would make many things much easier and open a new world of possibilities without actually resulting in more maintenance work for the ssg.

The thing is that the output is abstracted in that way, one can create a new plugin to write to a different output format. If one could also "generate" input/source files dynamically, one would get support for all of these output formats "for free".

As for me personally, I don't really have enough blog posts to have to store them in a db, that was just an example. But if abstracted transparently in the way I am thinking about it actually doesn't matter for the ssg, it only knows about inputs and outputs and not how it got these inputs.