They are not updated correctly, none of the individual results in the results files have more than one timestamp, even if the same relay have several results.
So there is a bug somewhere.
I could not spot it because currently we don't have a way to run whole measurement loops with an small number of relays (again #33150 (moved)).
I could add some debug logs and run it for some days, but since this is not very important right now and i found a fast work around, i implemented that as part of this ticket and create a new ticket to solve the real bug a some point.