I wonder if the spikes in my patch when considering snowflakes bridges are caused by this: https://gitlab.torproject.org/tpo/network-health/metrics/website/-/blob/userstats-bridge-count/src/main/sql/clients/init-userstats.sql#L290 I added the nickname at the end... https://www.postgresql.org/docs/current/queries-order.html says: When more than one expression is specified, the later values are used to sort rows that are equal according to the earlier value. And that comment says that the order is important in the merge function There is just one way to find out ;)