When a payrolls headline and the revisions disagree, which one do you anchor your read on?

A recurring puzzle in macro reading is what to do when the headline nonfarm payrolls number for a given month points one way and the prior-month revisions point the other. The headline is the first estimate and gets revised, sometimes substantially, so a strong print can look weaker once the two-month revision is folded in, and vice versa. The useful habit is to separate the headline from the detail: the level of payroll gains, the household survey, average hourly earnings and hours worked. Each tells you something different about labour demand, and they do not always agree in the same release. It also helps to check which vintage of data a given chart or commentary is using, since many charts quietly mix first prints with revised history. I'm curious how others here structure this in practice: do you weight the first print at all, or do you wait for the revised series before forming a view about the trend?