You're the inspiration for the current crop of 'clock-free' articles, I really got interested in it that particular aspect of our conversation after M's visit here.

I think I can see at least one way to crack that 'nut' of doing without a global clock, which would scale to arbitrary large systems.

The downside is that it is a purely theoretical solution for the moment, I'll try to do some simulations to see if it is actually viable.

The tricky bit seems to be to not fall in to the 'pit' of thinking that you have it working without synchronization when actually the test system has some underwater property that makes the likelihood of a failure vanishingly small, but which when duplicated 'in real life' would lead to an error almost immediately.

The classical example is two circuits that are 'independently' clocked from some master clock that gets divided by different factors.

You'd need to test this with local clocks that are varying wildly comparing the output of the system with 'known good' output over many cycles to gain some confidence.

It's amazing how the devil really is in the details with this stuff.