Hm, firstly I am very thankful for your contribution! While I feel a bit irritated that these issues seem to be known and discussed in disguise to an extend that people stopped using the i2c features whereas they still got promoted and further developed in the open it is somehow relieving to see that there are actually more people experiencing this and the discussion gets more open.
I love the monome system and really hope we can figure out a solution. Since the promised land of complex remote control was the deciding reason for me to get into all this and replace all my former sequencing/clocking/CV control modules I would happily contribute all I can to the process of exploration and debugging to eventually get the ecosystem running stable and reliable. I can't imagine playing live with it atm which is a shame cause I would love to!
Regarding the power context:
I tried teletype, ansible, earthsea, meadowphysics, switch and a grid in two different cases, one with a linear Doepfer PSU2, one with a Make Noise distro board and the external PSU they sell with their systems. None of both have been stable with i2c and I got lockups and periodical earthsea slowdowns. I even took out all other modules form both cases with no success. I don't know much about general electricity contexts but I've read it is pretty stable in germany - as you might think everything is statutorily regulated over here. For convenience and safety I run everything in my studio from a furman PS-8RII power sequencer.
The current system is extended with a second ansible/arc and sits in the case with the doepfer psu2 now together with Maths, Optomix, Twiggy and a Verbos line (CO, HO, ATC, multi-envelope). It is fully analogue besides the monome modules and uses about 890 mA max of 1200 mA. I have had two digital modules in that case that could not cope with switched power (which made me build the extra case with linear power) but taking them out made no noticable difference.
When I get lockups in minutes with the minimal freeze script I seem to use a faster trigger rate into teletype than on your video. But I had a slightly more complex patch (as described somewhere above) freeze after about 30 - 60 minutes that was very slow in every sense (M and trigger input) too. I assume that the probability of a crash just rises with faster rates.
If there is anything I could try over the easter days I have some time and would do it.
Could it be that I made a mistake building the new bus board? I would think it would not work at all then instead of being not much of a difference, right?