Here’s an interesting generative typeface, created by averaging a large collection of fonts on a computer: Avería.
Then it occurred to me: since my aim was to average a large number of fonts, perhaps it would be best to use a very simple process, and hope the results averaged out well over a large number of fonts. So, how about splitting each letter perimeter into lots of (say, 500) equally-spaced points, and just average between the corresponding positions of each, on each letter? It would be necessary to match up the points so they were about the same location in each letter, and then the process would be fairly simple.
The result is a surprisingly readable typeface, with an appealing hand-drawn quality.