I'm a little skeptical about a large number of multi-character chords, just because the frequencies of any n-gram are so low. In English, "th" seems worthwhile, and I also have "the", "an", "in", "and", and "you" but I'm not sure on the utility of them. I designed my layout years ago before I looked closely at n-gram frequencies versus individual letter frequencies. I probably wouldn't go beyond "th" if I were doing it today.
When you ask what the bottleneck is in typing, is it the physical ability of the fingers to move quickly enough, or is it the ability of the mind to plan sequences of key presses without hesitation? If it's the latter, then MCCs are only a net gain if they can be recalled faster than the equivalent number of keypresses, and if you have enough of them that a reasonable proportion of real-world text calls for MCCs. I don't feel that "in" occurs often enough to make much difference on its own, and that's the third most common bigram. There's a very long tail for n-grams. It seems enough chords to see a benefit would be beyond the point of practicality.
I'm interested in being proven wrong though. I wonder if some kind of steno system could be designed using the Twiddler.
