This could be a good allegory for sampling intervals. I’m forever having to explain to people that if they are looking at data every 60 seconds they can take a lot longer to notice a problem than common sense would suggest. Your new information shows up on average 30 seconds late and up to 59.99999 seconds late. If you’re trying to compare the last two samples to detect a problem then you’re going to see it up to 2 minutes after it already started, since it could start 1 ms after the first sample. Particularly in systems where multiple actors are using periodic actions.