You need to specify the process. We hadn't, just assumed. So analog and power boards were just no-clean flux. Only the guy doing 10 fA electrometer has the sense to specify cleaning.
No-clean is fine for 3.3V logic. And it makes the manufacturing process green, by distributing waste over landfills worldwide, rather than collecting flux residue at the point of PCB manufacturing.
But maybe it has nothing to do with your problem.
If you have spare channels, you could include self-test of surface leakage.
I've found some customer interface personnel to sometimes be compulsive liars. I asked how many parts in the lot were hand placed. he said none. I said to have no assembly defects requiring rework on our first batch of prototypes, they would have had to have a 10 dpmo process. He said they did. On a first prototype? A part was used 10 places on each board, 20 boards, and exactly one site was stuffed with a different tolerance part of similar value, that was used elsewhere on the BOM? (Turns out parts fell off the SMT machine, so the operator picked one up and hand placed it.) I wanted the board shop to give us a list of all defects spotted and corrected, because that's the first place I would look if a board malfunctioned.