Right...consider this: you have an LFO cascade with irregular but very long periods. You then set up a comparator to view this, then send the LFO curve on to a filter cutoff, perhaps with some attenuation. The comparator's gate output then goes to the start/stop of a sequencer, so that the sequencer only runs when the LFO cascade's output voltage exceeds X amount. Then you also send the comparator on to a XOR gate, which has a clock on its other input. When both signals are present, the gate sends nothing. But when only the clock is there, it passes and then goes on to some other device, such as a Euclidean sequencer's clock input, so that when one sequencer is running, the rhythmic pattern generation of the Euclidean stops, and when it's stopped, the rhythms pick back up again. This is what's neat about having all of these digital control subprocesses going on behind the scenes; they allow for an incredible amount of complex control staging that can be just as generatively controlled as pitch CV processes are in a generative compositional system. In short: it'll drive ya NUTZ...but that's a good thing!
The big changes in recent years, which partly (I think) got kicked off by Don Buchla via the early 200e modules, is the whole 'digital in analog's clothing' paradigm, where you have analog control over a module which, behind the panel, is based around some sort of DSP setup, either via a dedicated DSP, or a cheap micro such as an Arduino or Raspberry Pi. The ability to seamlessly merge that sort of digital generation architecture in with all of the analog stuff is blowing things wide open right now.