let's suppose that user A downvotes everything from user B, while most other users generally like the posts from user B. How likely is it that the community as a whole would benefit if the user B becomes discouraged by this behavior and leaves?
Unlikely.
Let's assume the user A behaves this way towards users B, C, D. In this case we have one person trying to send away three people, that other users don't mind. How likely is this to improve the website?
Unlikely.
Maybe it would be good to have some accepted way for the user A to express their dislike towards the user B,
Dislike is another matter entirely. What we're talking about is ways for A to express their preference that B not post here. And, as I've said, it seems we do have a way for A to express that preference: downvoting.
I agree with you completely that in the examples you list, and other similar examples where A's preference is a likely-mistaken one, any mechanism that allows A to effectively act on that preference will likely harm the site.
let the community decide -- a democratic ostracism vote, instead of an assassination. The key is that the community as a whole expresses their opinion, not just one individual removes another individual.
Sure, I endorse that.
For example, we could provide a mechanism whereby other users (E, F, G, etc.) can upvote contributions from users they consider valuable. Then the net karma score of users (B, C, D) would respect the collective opinions of the community as a whole, including but not limited to A's opinion.
The first situation that you call unlikely is empirically happening. See Daenery's comment here. The second situation you call unlikely also seems to be happening given that multiple users have reported the block downvoting to be occurring in a very similar fashion, and the political motivation in many of the cases looks identical.
I'm just tired of the signal pollution, and would like to be able to use karma to honestly appraise the worth of my articles and posts, without seeing 80% of my downvotes come in chunks that correspond precisely to how many posts I've made since the last massive downvote spree.
EDIT to add data points:
Spurious downvoting stopped soon after I named a particular individual (not ALL downvoting stopped, but the downvotes I got all seemed on-the-level.)
One block of potentially spurious downvoting occurred approximately one week ago, but then karma patterns returned to expected levels. I consider this block dubious, because it reasonably matches what I'd expect to see if someone noticed several of my posts together and disagreed with all of them, and did not match the usual pattern of starting with the earliest or latest post that I had made and downvoting everything (it downvoted all posts in a few threads, but not in other threads), so I'm just adding for completeness.
Spurious, indiscriminate downvoting started up again approximately half an hour ago on Sunday (12/1/2013), around noon MDT.
Edit: And now on Tuesday, 12/3/2013, at 10 AM, I'm watching my karma go down again... about 30 points so far.
Edit: And now on Saturday, 12/14/2013, at 2 PM, I'm watching my karma go down again... about 15 points so far, at a rate of about 1-2 points per second.