I got another epiphany that in retrospect seems out to be terribly obvious – the second half of a diphthong doesn’t have to be very long.
For example, the OY in BOY is a diphthong, which means that the vowel has two targets. So the question is, how long should the YX target be?
For singing, the answer is “Not very much” – especially in the case of a YX sound, which is rather harsh.
The really sad bit is that I’d already discovered and implemented it some time ago… And then forgotten about it when I recoded the core.