Interestingly, although it seems harder, this sort of analysis is much more reliable with proportional fonts, because the specific width of the gap narrows the number of possibilities down considerably.

That would would depend heavily both on the proportional-width font that was used and the accuracy with which we could measure the width of the gap.

You don't really even need to measure widths, since they fall in a grid. Just count the number of characters directly above or below the redacted passage. (For fixed-width, I mean.)