There was a race between the background WTOplogJournalThread and the test running inserts. The inserts set the WT transaction timestamp to Timestamp(1, 1) and that was why query_timestamp returned Timestamp(1, 0). SERVER-42497 added the setTimestamp call to the inserts.

In production we don't have any logic that sets timestamps to anything other than what the LogicalClock timestamp generator gives us, so this race cannot occur in production. Ideally, no dbtest would explicitly set the timestamps backwards like this.