On an ideal world, one would have all the time one needs to try out each and every reply to see which was best.

Why ask a question in a public forum if you're going to ignore some of the answers?

... the earlier replies tend to get more votes, but replies that are technically wrong do tend to much lower (to the point often of going negative) reputation.

If the strongest certainty you can offer as to the correlation of reputation with correctness is a tendency, what does showing the reputation of a node after a null vote give you that ordering by reputation doesn't?

Again, to see the reputation of replies, the questioner would have to cast a null vote on each reply. At low levels, he won't have enough votes in one day to vote on every node attached to many questions. With reply ordering, he doesn't have to spend any votes. Sure, he can only see a rough relationship of one reply to the other, but what does the fact that one reply has a reputation of 18 and another has a reputation of 12 explain?

You can't expand one bit of information (+1 or -1) to "technically correct" and "technically incorrect" because that bit has already been compressed from myriad reasons. "I don't like this node." "I don't like the poster." "I disagree, but it's a good thought." "My finger slipped." "I'm trying to use up my last vote." "The poster needs one point to go up a level." "I think he downvoted me elsewhere."

This proposal strikes me as trying to squeeze way too much meaning out of a number that's mostly meaningless already, especially when a solution already exists without the drawbacks.

I'm all for making reply ordering by reputation the default, though.