Thanks for the great article and crediting Caleydo LineUp. Last year the Caleydo team published the work about CLUE (Capture, Label, Understand, Explain), wich is similar to your work with redux-dag-history. You can take a look at the research paper and demos at http://caleydo.org/publications/2016_eurovis_clue/ .
> For example, there are non-trivial memory implications about preserving multiple copies of application state without limit.
To avoid (or workaround) the provenance graph in CLUE (in your case dag-history) can be either stored in the browser’s SessionStorage or remote on a server. But this is also not the perfect solution.
One important difference between both concepts is the state itself. A redux state contains always the full application state and the application can be restored with a single state represenation.
However, in CLUE all subsequent actions lead to a state (similar to a difference backup). This means for restoring a state all actions that lead to this state must be loaded and executed.