Query sampling-corrected volume
Multiply stored counters by each row's sampling rate, against raw flows or a rollup, with the boundary predicates a rollup needs and the checks that tell you whether they apply.
Tune queries over samples()
Keep a query over the samples() view fast by preferring a rollup at minute buckets, bounding both the flow interval and the raw record time, and rendering labels after aggregating.
Backfill a rollup
Populate a 1-minute rollup from historical flows rows with an INSERT INTO … SELECT that names every column in the view's own order, and verify nothing was shifted or omitted.