Indeed InfluxDB does not sort on anything but time (in that case I believe SLIMIT would do what we’re looking for).
But today, top() does the trick anyway. I have a point for each user every interval, like
An InfluxDB query does that with
select top(cpu,3),user from usage where time > now() - 30m
But I can’t see how to convince Grafana to use the user from the fields rather than from multiple result sets like you get with GROUP BY.