>
> Opinion: if a soft work works, it should be preferred, if for no other
> reason than once a hard-fork is planned, the discussion begins about
> what else to throw in.  To minimize the frequency of hard-forks, the
> time for that is when the change being considered actually requires one.


I'm not sure why it'd be any different. Soft forks are just as disruptive -
everyone who needs a correct node has to upgrade on time. Given that, I
guess there will be a desire to roll out several changes at once too,
regardless of what happens to older nodes.