Unexpected empty result set for alert?

I’ve read the Alerting docs a couple times, and it’s not 100% clear to me what time range the alerting rules are supposed to operate on. Time range being the “query(A, start, end)” part of the conditions.
Is it the clock timestamps when alerting checks are performed?
Or is it the time variable constructed by the query itself? I had assumed it was the latter, but I’m not sure with this new alert I’m trying to construct.

Here’s a working alert’s test output: https://i.imgur.com/XfOzPuil.png
There’s an array for the series output, and interestingly, rowCount = 0 for the sql output.

Here’s the output for my not working alert: https://i.imgur.com/awdYZOV.png
So the series is null, and yet the rowCount for the sql is 8.

Here’s the query and the 8 rows:

=> select date_trunc('week', valuntil) as time, count(*) as ct from pg_user group by 1 order by 1;
          time          | ct  
 2019-10-21 00:00:00+00 |   9
 2019-10-28 00:00:00+00 |  55
 2019-11-04 00:00:00+00 |  96
 2019-11-11 00:00:00+00 | 141
 2019-11-18 00:00:00+00 | 274
 2019-11-25 00:00:00+00 |  31
 infinity               |   3
                        |   3
(8 rows)

The time range I’m specifying:
query (A, 240h, now)

Which should capture 10 days, right? Picking up at least 2 rows, if it’s filtering off the time column.
It’s not apparent to me what Grafana is doing with the 8 rows that it can get back from the sql.