The reason no one does multitimbrality anymore is that people making music is getting used to recording audio with their synths... they like to be able to record as many tracks as they want with audio from the same synth, thus needing less physical synths... i bet this is why multitimbrality is phasing out with many companies.
The main reason for some being dual or even a third layer is that performing musicians need stacks/splits and thus more than 3 would be more or less useless. multimode in many of these synths is a simple bonus where you can control the layers independently via their own MIDI channel, and code wise it's pretty easy to include when you've already got stack/split functionality.
A third reason may be that any synth with analog parts in the signal chain is always limited to about 8-16 voices ... if you had 4 parts on a REV2 16-voice you'd be down to 4 voices per part... we're reaching a polyphony that's becoming insufficient for anything but leads, basses and other short release sounds now, and when you use multitimbral you all know that the first question will be "does it have separate outs!?" which further increase the cost of the synth, and complicates any built in FX that will usually be routed to only one output, otherwise the FX engine have to be duplicated for each output as well, further increasing cost.
With a fully digital synth, it's much more feasible to have multitimbrality, especially these days with the horsepower the CPU's give you... probably also why the KYRA have 8 part multitimbrality... it also has 128 voices giving you 16 voices per part because of this.