Cumbersome way to calculate temporal difference between two cubes from custom processes

# load before flash flood params
before_date = ["2021-06-13","2021-06-13"]
after_date = ["2021-07-18", "2021-07-18"]

# Create a processing graph from the NDWI process using an active openEO connection
before_ndwi = eoconn.datacube_from_process("NDWI", namespace="vito", date=before_date
                                        ,polygon=aoi)
# Create a processing graph from the NDWI process using an active openEO connection
after_ndwi = eoconn.datacube_from_process("NDWI", namespace="vito", date=after_date
                                        ,polygon=aoi)


from openeo.metadata import CollectionMetadata

before_ndwi.metadata = CollectionMetadata({"cube:dimensions":{"t":{"type":"temporal"}}})
after_ndwi.metadata = CollectionMetadata({"cube:dimensions":{"t":{"type":"temporal"}}})
# final result
# compute the change between pre and post image
merging_cubes = after_ndwi.merge_cubes(-before_ndwi)
differenced_cube = merging_cubes.reduce_dimension(dimension="t",reducer='sum')

#differenced_cube.download()

Start by reviewing the shown datacube_from_process calls and the merge_cubes and reduce_dimension entry points. Clarify the intended API for calculating a temporal difference between the two custom-process cubes, then define completion criteria and corresponding tests before changing behavior.