You are slashing your way into the Papa New Guinean jungle...

This isn’t it.

Historical models "were" a beginning in the past. Yes, but they need not be "The" beginning for today, unless what you are trying to do is replicate them or revise them. The point of drawing a new typeface is drawing a new typeface.  Don't worry about would be critics.  Use your eyes and mind to define what makes a letterform concept into a finished glyph, and more importantly, what makes all of the glyphs work together as a family unit.  You are not looking for Jensen's answer, you are looking for your own thesis. If you look at World scripts, you see variation in approach.  None of them are wrong, they are just what emerged over the centuries. There is no recipe that guarantees success, only a meager place to start.  Judge what you see in front of you as you make it and then put it away and judge again--you will see more. What you need to do is do the work of teaching yourself to see, not looking for the Holy Grail. Put in the work, then you have something to judge.

To make a less objectionable “pre-contact” comparison, Craig, consider the ethnological group of kids today who have grown up with sans faces on mobile devices, and have never read a book printed in Bembo or Garamond.

or else, just start drawing type for a good chunk of time

I wouldn’t say studying historical models directly is necessary (although it’s useful and one possible approach). I think having a deeply internalized understanding of what the reading public expects is essential. From there, one can choose to follow or break the rules, but will always see and understand what they’re making against that baseline.

And I certainly agree that looking at one’s own typeface in text samples is necessary, and really putting what I wrote above into practice. Looking at individual letters or an alphabet out of context is not particularly helpful (and often counterproductive, I think), while looking at text settings is literally seeing one’s typeface at work. A typeface is a system to make words visible and readable, so if you’re not looking at words, you’re not really looking at your typeface.

The trick is to look at a word, a sentence or a paragraph and see what doesn’t work. One can’t do that unless one first understands what does work. Given that what we know that works can be traced back to historical models, one can either go back to those sources or just look at successful, recent examples that have already benefited from those historical models.

(I’m talking mostly about text typefaces here, although display faces aren’t so different.)

Proportion is a big part of this, because that’s about the relationship of one letter to others. Getting proportions right is about finding harmony among the letters that are put together to make nice text. As others have said, there’s a sweet spot along the line from monotony, liveliness, to dissonance. One has to follow and break the rules enough to convey the right amount of functionality and pleasure without breaking into territory that becomes irritating — and there’s no better way to find that spot than by simply looking, but looking won’t work if one isn’t able to spot what’s unfamiliar or wrong, and understand why.

To practice and learn all this, I think revivals and imitations are great. The world has plenty of Garamond revivals, but making one (not necessarily to sell it) will teach someone a lot before they understand how to create something new and interesting. It’s the equivalent of spending years in a cover band playing standards. It worked for the Beatles, and it’s how one becomes an accomplished artist by first mastering the mechanics of their craft.