I use the wait time chart as a general guide to ongoing performance but I have noticed that the early production hours have a habit of changing during the day, long after the hour is finished. I know that this would happen if I change the excluded query list (although I haven't done so) and that it might occasionally happen because of a long running query, but it appears to be mainly due to a design problem.
My suspicion is that this is because DPA is only displaying the top X (or Y%) of SQL Statements and so from 8am to 9am the statements are typically much shorter than the ones during the overnight processing. It is only later in the day, maybe 11am or noon, that those statements take enough proportion of the total activity for them to be displayed.
The solution would be to display either an "other" section or to display an indicator showing the "Total" wait time for each period.
The result of this is that, at 9:05am it looks like the total wait time during the 8-9am hour is maybe 100 minutes while, at 10:30am, it is 400 minutes... Meanwhile, the wait times from 2am to 4am have actually declined;
A similar effect can happen from one day to the next, although it is usually less obvious. For example, for the 17th October I get these two different charts - one from the next day and one from a week later - the main difference is that the latter one includes SQL Hash 4592265194 between 5255831054 and 3411566237 (note that the scales are different, but the cumulative wait time shown on the latter chart is still longer than the first chart even when the scales are adjusted);
