Currently, every tweet ID returned from the search is queued. these are then checked 1:1 on the queue listener named VisionScanner.cs
Instead, what would it look like if the search pulled all the tweet id's, bulk queried storage, and only queued new items?