OK I wont go into full details, but one of the major problems is qualifying the powder used in the 3d printing process.
Maybe I should go back a step. In the most common type of Metal 3D printing, called ALM, you lay down a very thin layer of Aluminium (or Titanium, etc.) powder, a laser then melts the powder in the specific areas you want, fusing it to the layer below. You then drop the level a tiny bit, roll on a new thin layer of powder, and go again with the laser. You keep doing this until you're finished. Then you suck up all the excess powder you can, for reuse. Then you dig out your part, clean it up, remove the support structures, and you have your final part.
In a billet of solid material, the characteristics, impurities, etc. spread through the metal in a surprisingly even way, so testing one part of the metal is analogous to testing any part of the material (or at least its considered good enough in modern systems). In powder, that's a much harder prospect to justify, so it's a lot harder to justify that the powder in the left hand corner of your machine bed is exactly the same as that in the right hand side. Or that today's powder will have exactly the same properties in the same places as tomorrow's part. You also have a major problem with proving that the material currently in your machine is identical to the next batch of powder you refill with, even if it comes from the same supplier, etc. Imagine how much more complex that problem becomes, when you consider that after a half dozen refillings, you will have a powder that is actually a mix of all 6 powders, each with their own minute differences in chemical properties. It means you cant just use the CoC's of the latest powders, to justify the quality of your powder. On top of that, at least some of the powder you collect for reuse will have been heated by the laser, not enough to melt and affix to the desired structure, but to mildly change the properties. You cannot filter that out, so you have to account for that in your qualification, that perhaps you get all of the "bad" powder in a critical position. Because of that sort of thing you also need to have higher safety factors in your stress analysis in order to account for the lower confidence in the powder.
One final thing, your Stressies also need to account for the fact that unlike a billet of solid material, where properties are isotropic (the same in every direction), ALM parts are Orthotropic, they are the same in the plane of the powder bed, but the through direction is significantly different. And because you normally, optimise your orientation in the ALM machine to minimise problem manufacturing areas (support structures, hole directions, sharp edges, obtuse angles, etc.), your desired principal strength axes, almost never align with your actual manufacturing axes. So 3D printed parts are a right pain for stress engineering. (Thankfully I'm a designer, so I make the issues, I dont have to solve them :P). Our firm, basically creates mini stress-strain test samples in various orientations around the manufactured part, so that you can at least test what the strength at 45° in the North-West direction should be in your actual part, without having to do some sort of destructive test.
All this makes it a very long process to get something qualified for flight, and you tend to be heavier than optimal (because of the safety factors). Done right you can still come out lighter and cheaper than a standard manufactured part/assembly (when you take into account things like being able to design a single part to replace multiple parts and remove multiple assembly steps), but it's certainly not an easy process. Admittedly, if you dont have the major qualification requirements of the Aerospace industry, then integrating ALM printing into another industry can be a LOT easier, faster and likely cheaper. But keep in mind that ALM (and 3d printing in general) is best suited to low volume, high cost items. It's actually extremely good at replacing moulded/cast parts. But keep in mind that ALM is totally not relevant for high volume production.
Sorry, that went on a bit longer than I planned. ;) But I hope it gives you an idea of the problems. Hence, I'd love to hear how NASA are sorting the qualification side of things, so if the Author ever reads this, would you consider doing a follow up article... ;) :P