After doing much thinking and learning from experience with recent dry runs I have concluded that we need to have negative votes.
A negative vote is morally equivalent to a vote of approval for "everyone but X" like a positive vote is the moral equivalent as a vote of disapproval for everyone but Y.
In a world where the squeaky wheal gets the grease, it is far more likely that the average user will know more about who they are against than who they are for. When there is a bad actor it is far more difficult to get people to up-vote 101 other candidates than down vote the guy causing the problem. For automated voting it is also far easier to down vote delegates that are not producing blocks than to decide who to up vote.
The average user is lazy and unlikely to evaluate 101 different individuals. We must factor in this laziness in our designs.
Lastly in a world with 1000 candidates, it takes a lot of work to concentrate the approval consensus. Suppose Bad Actor has 25% approval, everyone else has to find someone they can agree to give 26% approval to bump the Bad Actor.... there are 2 solid candidates and they each split the vote 13% / 13% and thus neither is able to bump the Bad Actor. Everyone can agree the Bad Actor should go, so they simply vote against him.
Negative voting is effectively an easy way to compress a "vote for all, but 1" which in a way helps simulate the ability to approve more than 101 delegates.