Not that someone somewhere hasn't already done it, but a potential problem with that approach that immediately comes to mind is you need the image of mask itself to be diffuse as to be unnoticeable but not so diffuse that it causes glowing halos around edges. Another is that you'd also need the mask to be precisely the same size as the final copy, the registration issues now being dimensional, as well as mere alignment. Registration is relatively simple to do with large format when you contact a neg or transparency but now that's going to present something a challenge on the easel epecially if you'll be enlarging (Might find you'd need a perfectly flat field enlarging lens, too).
None of which may be such a problem in the larger formats contact printing at 1:1, though you still might need watch out for introducing an interference pattern (moire) with the slightest mis-alignment. Film masks would be relatively immune to this, though they do bump up the visual grain a bit. (Again, a diffuse light source helps keep this manageable).
Add to all this that if starting from a smaller format than 4x5, you're probably going to need a dedicated film scanner of the highest resolution in order to make contrast masks that can withstand any degree of enlargement without pixelization being obvious. Which begins to beg the question, if you're going to all that trouble, why not just hand off a masked scanned file to someone with a Chromira who runs Ilfochrome through it? Speaking only for myself, the whole point of any masking exercise is to improve upon the image, not detract from it-- which to me means it certainly isn't worth the bother and additional expense just for bragging rights or some gee-whiz factor of having hybridized the process.