1. distinguishing multiple metrics for comment evaluation, and
2. declaring that up/down voting should not be used for some of these metrics
are doomed to failure, so long as
1. there is only one dimension on which to express anything, and
2. the interface for voting lacks any obvious affordances/constraints on how it should be used, even if those affordances/constraints are merely rhetorical (e.g., a JavaScript popup that asks you to confirm your up/down vote only if complies with some set of criteria).
I suspect PG is unlikely to complicate the voting interface to map to multiple metrics until some threshold of community deterioration has been reached, so a lot of this debate is pretty pointless. The only way to enforce norms about how generic up/down voting should be used, that are not implied by the interface itself, is to constantly comment on uses of voting that are suspected to be illegitimate and to start threads like this one.
After seeing several such threads, I suspect that, at best, you will only be able to convince a fraction of the community to conform to whatever norms you think are best. Everyone else will simply use the up/down arrows as a catchall for any positive or negative sentiment they feel like expressing. Total conversion is pretty much impossible, partly because some people will disagree with those norms, and newcomers will be oblivious to them.
In my opinion, there are much bigger problems here than how up/down votes are used. Constant submission and upvoting of news that's utterly irrelevant to hacker-entrepreneurs is one.