They succeed for me (with multiple runs) on x86_64-darwin native (no
Rosetta 2). Afaik, Hydra performs x86_64-darwin builds on Rosetta 2 on
aarch64-darwin. I do not have the hardware to reproduce this (and it
could be a build issue), so let’s disable the tests for now and go from
there.
Unfortunately one of the issues is within a dependency, so have to do
some awkward patching of the go-modules. This patch is pending merge to
the upstream dep and will hopefully eventually trickle down. We do *not*
optionally apply the patch for Darwin only because then we would have a
different hash per system, which seems even more awkward.
