"database vs. text" is a false dichotomy
Clearly we disagree.
It is a dichotomy - the programs either work using a database or they use files. (Or potentially both in the case of WriteMonkey 3).
"data ownership" (not a given with a DB, e.g. if it's using an open DB format and/or all content is markdown and it has a robust exporter).
Here you are introducing caveats specifying things to be watched for in a database.
Such watching is unnecessary with files.
In practice, databases don't use pure markdown (although they may attempt to export it ). The file based programs aren't pure either as they all prefer wikilinks. Not that any markdown is really pure.
And exports from programs like Roam need a lot of work to convert to useful files - and that is likely to increase as Roam continues to add features.
In theory, a database can provide any feature that a program using files has.
There are additional benefits, such as only having one file to backup. And speed - everything is loaded in a usable format whereas files have to be loaded before any processing.
But also disbenefits - they're rigid, you can only do what the program does; databases corrupt; export may not be as robust as you thought, especially if development is rapid.
The differences make the difference between the approaches a useful dichotomy to consider. For many, maybe even the majority, the benefits tip on the side of the database. Going with either approach means accepting the disadvantages of that choice.