Currently TROPESS downloads entire sets of L1B data from GES-DISC for calendar days at a time. I then uses those files to extract meta data that is used to thin, pair and batch soundings down to a smaller subset that will be used in the the retrieval.
Now here's an idea, what if we just did all those steps with the metadata obtained through API calls (earthaccess) without reading the actual file themselves.
That would save so much time.