I definitely should have added that one, though. It's in the list now.
It's definitely an "old-school" construction, similar to what heisenburp.py shows -- use up the input, send a fast signal to get ahead of where the input would have gone, and put an input back there.
Guam's LWSS pseudo-Heisenburp design from back in 2011 was way ahead of its time -- or right at the beginning of its time, anyway. The output LWSS shows up almost instantly by comparison to the signal wire catch-up method.
Maybe it would be worth running all the octohash constellations through collisions with all possible *WSSes ... see how many clean *WSSes come out, and if any of them are cheaper than Guam's contribution.
(It could be two different *WSS->*WSS one-time converters, as long as one retimed / shifted the output in the opposite direction and amount as the other.)
Guam's constellation only needs
four synchronized gliders from two directions, but I haven't checked if Guam's suggested HF+loaf -> 2G converter is compatible with it. Would have to rebuild the loaf, too, in that case.