Saw a Harwin D-sub datasheet that legitimately and utterly screwed up the drawings, completely wrong view, or pin 1 on the wrong side, something like that, I forget. Fortunately the customer noticed something fishy on review!
It's never done until the part and calipers are in your hands. Until then it's just guesswork!
I think it's really amusing (somewhat in a "well, f*ck me" sort of way) that most manufacturers give out STEP files, that are almost always more accurate and representative than their very own datasheets, but that they always disclaim them as "for reference only" and such.
I really should insist on pre-buying samples of everything I work with, but it seems such an unnecessary step for the most part. And D-subs, those things have been around for almost a century, MUST be fine, right?
The other part I'm fond of shaming, is a cell holder (CR2032 or such, SMT) by MPD. Only the side-of-barn-sized global tolerance is given, which if I were to follow properly for all position and diameter dimensions as labeled, would require holes for the alignment pegs which are big enough to fit either peg (plus enough positional and rotational error to just about fully miss one or both pads). It was something like, an 0.8mm and a 1.0mm peg to set polarity and alignment -- the tolerance was 0.25mm, applied to not just the diameters, but the spacing as well.
Tim