I'm trying to defend Holman here - the Leach paper is from 1987, almost a decade after Linkwitz's original paper. The ideas that Leach shared in that AES engineering reports was one of those things that probably popped into the heads of many people who read Linkwitz's original paper right then and there. If you're an EE with a background in filter design it might be pretty obvious. I know that we had implemented our first L-R crossovers that way a year or two after Linkwitz's original paper, not thinking that it was rocket science.
If memory serves we were looking at the effects of driver roll-offs on measured response and asked ourselves the question "Would this be smoother if we bypassed one of the two stages of Butterworth filtering". It was. If we could figure it out, a designer of Holman's stature would probably get it all by himself even sooner.
So if you know the (ideal) -3dB frequency of the mains, and they are sealed and have QTC near 0.707, and you have the ability to set the crossover frequency to the -3dB frequency of the mains, you can generalize the THX approach.
Exactly. Caveat - the mains should be sealed boxes, not vented. Venting gets you 4th order Butterworth which is a shade different from the cascaded second order Butterworth filters that Linkwitz specified. Also not everybody tunes their sealed boxes for maximally flat Butterworth response. Many speakers are a little underdamped for warmth.
Let's be real - the acoustical response of speakers is not nearly as exact as electrical networks. Everything we do is observed through the blurred lenses of room acoustics and human perception.
One thing I like about the Onkyo receivers is that they have the biggest choice of crossover frequencies, making this approach more practical.
But 40 Hz seems rather low to try this, because the electroacoustic phase shift of the sub probably hasn't gotten close to zero at this frequency yet.
When you're doing an acoustical/electrical Linkwitz Riley, the phase shift of the sub is part of the design and you actually hope that the electroacoustic phase shift is there to match the design specification. That's part of the charm of the approach - from the standpoint of the design the normal electroacoustic phase shift of the speakers can actually work for you to achieve a higher goal.
Also, lets not forget that the final response of a properly-implemented Linkwitz Riley crossover is a big fat all-pass filter with a quick 360 degree phase shift at the crossover frequency. What it does to square waves isn't pretty at all. It even takes an observable toll on tone bursts.
However, have no illusions you are choosing the lesser of two evils. Changes in speaker directiivity though the crossover frequency can be very audible because it can cause frequency response variations that might not be expected if you just look at the magnitude response of the two speakers being blended, and even if you factor in the magnitude response of the electrical crossovers.
I can't think of any musical instrument that puts out clean tone bursts or even far less probable clean square waves. A little ringing and some group delay is the lesser evil. We observe these effects through the very fuzzy lens of room acoustics and the somewhat fuzzy lens of our hearing. Anybody with illusions about the sensitivity and reliability of the ears as test equipment take notice!