[cfallin]

I did something like this when I was doing lots of data analysis for a research project: [1]

The problem space I was addressing was more restricted than a fully generic spreadsheet model, but I found everything you say about reproducibility and descriptiveness to be true. I kept my report/plot descriptions in git alongside the experiment setups / parameters.

FWIW, prior to this, I often wrote one-off Python scripts that would generate, e.g., 1000 rows of a particular set of formulae in CSV then import that into Excel. It probably wouldn't be hard to build a little tool in a few hundred lines of code to do this.

[1] https://github.com/cfallin/scomp