For the dumb relay board I'm thinking there are variables in the path from input trigger to relay contact that might favor CH1 in your particular board. There is an opto-coupler which have notorious part-to-part CTR variations. This then cascades into the relay coil drive current which depends on the beta of the NPN buffer. Then the part-to-part variation in the electro-mechanical response times in contact release/close. I'd guess this latter is the dominant thumb-on-the-scale; in other words if just the blue cube relays were swapped between CH1 and CH2, then CH2 would "take over."
I'd also think since the trigger currents in these dumb modules are about 1 mA, that a 5-cent R-C filter could be installed on one feedback path to handicap the race. I'm just gabbing in the spirit of a discussion forum. In the real world of OGR, just the pair of 1-cent diodes can be a show-stopper!
