We could store them in a SQLite database or in an HDF5 file. Two outputs are missing for risk, they are being computed by @raoanirudh in postprocessing and should be added to the engine: AAL aggregated by taxonomy and by macrotaxonomy.
We can also import in the workflow file the hazard outputs: for that, it is necessary to add a new hmaps-stats output consistent with the hcurves-stats output (i.e. with a CSV file per period).