What is the right way to read out a collection?

After a long dataflow computation, I end up with a collection that contains some answers to some questions I have. (Perhaps, using database terminology, the collection is the output of a query that I'm interested in.)

At some point, I want to pass the contents of this collection to some non-differential code (imagine, outputting the collection as the response to a REST query).

I don't see any guidance in the documentation on how to do this, but the two following ideas come to mind:

1. Use inspect directly on the collection, and build something entirely outside of the dataflow which is responsible for taking those diffs and maintaining a Vec or similar. However, this seems fairly heavyweight.

2. Use differential dataflow to create a new collection which just groups the collection in question into one giant Vec. Then inspects on that would just give the direct answer. However, this feels like it might be inefficient?

I'd love a pointer from you on what the best (lightweight, efficient) way is to do this operation! I hope I didn't miss anything in the documentation! Thanks!

The issue names no files, tests, or entry points. Start by reviewing the existing documentation and the collection, inspect, and Vec usage described in the question. Done means documenting a clear, lightweight recommendation for passing collection contents to non-dataflow code, including relevant trade-offs.