criterium problem as my suggestion, except for a different voting system:
> Do a rank-balloting among all of the options, with D as one of the options..
> Do a Schwartz Woodall count.among all the options.
> If D wins, or if the winner loses pairwise to D, or if the wnner
> doesn't beat D by the required supermajority ratio, or if number of
> voters doesn't meet the global voting quorum, then have further
> discussion, and later vote again.
> Otherwise, declare the winner as the chosen option.
There is also a certain Kurt Roeckx saying that this is not a problem: :)
> I think we do theoretically have this problem, and maybe we should
> change. But I think that practically we don't have this problem,
> and so I think we have little motivation to change it.
While it is true that it didn't end up affecting the init system vote, it very
clearly was something that could easily happen, either intentionally and
unintentionally (unintentionally if the upstart people genuinely preferred
FD to D in my example).
I don't know enough about Michael Ossipoff's suggested complete
change of voting system to have an opinion about that. But I think
that if Debian stays with the current Condorcet method, then my
(and Michael Ossipoff's I guess) suggestion of moving §A.6.3 down
as the last step is a simple and obvious must-have fix.
Looking at Wikipedia, I don't think the "official" Condorcet method
has Debian's default criteria. So presumably the §A.6.3 default
criteria's placement in the Debian constitution was not based on
well-reviewed research, and it is therefore not unreasonable to
assume that we can make it better by such a trivial tweak as
to move it down as the last step, as I suggested.