Your request is clear and the posts touch on some of the elements, let's see if I can relate it to an endeavor that DIDN'T work too well (we never learn from things that go great) and wrap it up for you.
The problem at hand (complex shape) used to be handled with a "loft drawing", basically the contours at various stations, like topographical maps showing lines of constant elevation. These were used for cowlings, canopies, fairings, etc. and the parts were made on a tool. Of course the drawing says "make on tool xxx," but what to do in case the tool wears out. Or how to make the tool in the first place. The loft drawing was also used to make templates which can be used as go/no-go gages. Typically fine for the tolerances involved.
Nowadays, coordinate measuring machines like a Faro arm can be used.
The program mentioned was more ambitious than just 3D CAD, one of those "change the world" efforts. One early shortcoming was excluding certification personnel in the (thought) process. The powers that be had the same PowerPoint presentation, so they thought there was nothing else necessary to satisfy the FAA. Never mind one thing that has NOT changed is the definition of Type Design data, information "necessary to define the configuration and the design features of the product" and "information on dimensions, materials, and processes necessary to define the structural strength of the product". So tell me what a purple blob on a screen is supposed to signify.
How this all panned out was that everyone was going to get viewers, FAA included, but like your feeling is telling you, that was not enough. Basically, nothing conformable, no STC. An early suggestion (not adopted) was to include a few key dimensions, i.e., mating interfaces, so the DAR had SOMETHING to write down, and the stress analyst could at least recognize what needed to be done (not much anyway - primarily passenger convenience items). There's nothing that says you can't have a set of data for certification, and another set for manufacturing. Most companies do anyway, in the form of process routing sheets, travelers, work cards. Only trick is that these cannot be different than the certification data (thus the Engineering Procedures Manual!).
So, a few inspectors got trained in 3D CAD and running CMM's. Imagine having to teach SolidWorks to your DAR.
Advice would be to think of your downstream users. Your machinist only needs a toolpath, so he does not need a "drawing". Your DAR and DER do, so spend half an hour and throw a few key dimensions on whatever you eject out of the printer. Even better yet, talk to the stress analyst and see what he needs. You might be shocked on how little it actually is. Many dimensions and specifications are for planning purposes, not getting it certified.
As far as your first question regarding the Engineering Procedures Manual, yes, it's useful for things like how you ensure data integrity (meaning revision control). But every time you expect some new type of document to solve a problem, all you may have done is avoid the original issue. Do it only if it makes sense for the organization as a whole.
