do randomly-selected sampling of paper buckets. e.g. you transmit your digital results, on a per-machine level, at t+1 minute to the central precinct. at t+5 m

do randomly-selected sampling of paper buckets. e.g. you transmit your digital results, on a per-machine level, at t+1 minute to the central precinct. at t+5 minutes the central precinct sends back a random request to have X% of the machines audited by hand. If there is a significant abnormality, we move to a larger recount and withhold electronic results. If things match up, we announce the full electronic results around t+1 hour. Works for me?