We've discussed the TERRA-REF hyperspectral workflow as a possible example:
In the case of the hyperspectral extractor development, the hyperspectral data is too large to move so the researcher/developer working on related algorithms needs to work on data in place. This is done via a NetCDF/NCO-specific development environment, but could just as easily by Jupyter or RStudio. If the researchers needs exceed container constraints, they can apply for dedicated cloud resources via OpenStack. They also have access to run jobs on ROGER via PBS. This is similar to Cyverse.