Thanks, Rene. I am learning.
This almost certainly means all 6 bugs I raised at the same time are wrong.
It also indicates a problem in the automated build-to-repo pipeline. Packages which are 'Multi-Arch: same' maybe should not be exposed in the package repository until all arches have built (and yes I know that's a whole nother ball of worms, it would need heuristics to time out and expose them anyway if some build was stuck for a long time; maybe only withhold an arch if another arch was pending which was 'known to co-reside a lot' like [ amd64, i386 ] or [ arm64, armhf ]...)