So it turns out that the cause was indeed a rogue change they couldn’t roll back as we had been speculating.

Weird that whatever this issue is didn’t occur in their test environment before they deployed into Production. I wonder why that is.