Move away from Spark JobServer to an Akka HTTP Service that runs newer GeoTrellis and takes advantage of the Collections API for a significant performance improvement of the order of a magnitude.