All interactive plots are created via the Vega-Lite or Vega API. Both the derivative datasets and visualization specifications can be auto-generated via the python script that is manually run before a PR is submitted.
In order for a Vega file to be rendered it will need to refer to a dataset. For most plots, the base dataset files will not be formatted appropriately even though all information is present within it.
Currently, derivative datasets that are used for plots are generated in the python script mentioned above.
If the python script is altered, it will need to be executed before any PR is accepted, so that all derivative files are updated before being hosted. This process will eventually be automated (hopefully).
Alternatively, Vega and Vega-Lite have a transform grammar that allows for data transformations in the specification itself. This is more complicated but cleaner way to provide visualizations.
If this is used, then it is recommended to use the large holders_munged.csv dataset as it has all base data combined into a single set.