Also the flow storage is fully plugable (https://nodered.org/docs/api/storage/) so nothing to stop you choosing/implementing a different storage format e.g. YAML https://github.com/natcl/node-red-contrib-yaml-storage
It's an array of objects, each representing a 'thing' in the flow. Each object has a unqiue id (which can be any randomly generated value), and other meta data that defines what the object is and what its configuration is.
Given most nodes have been implemented with a graphical UI in mind for configuring the nodes, you'd have to refer to individual nodes to understand what node-specific properties they require.
Maybe I'm not/HN isn't 'the community', but when I mentioned it 6 months ago [0] you said it was 'an area [you'd] like to improve'. ;)
The relative popularity of my comment above suggests others would be interested too. I understand it's not the focus/target audience necessarily though, I was clearer about that in [0], just a shame (for me) that the low/no-code isn't built on top of code(ish)-allowed.
It could have been an integral part (and really useful!) of ETL pipelines at work, but not without a git-sane, text-editor-editable file format. (JSON ok - not the best perhaps but ok - I mean more about requirements of IDs and general structure of the document.)