I think it also caused bad behavior. I found myself not wanting to reply to people's comments if the post was old, as my comment would be less likely to be upvoted, and my average score would go down.
This is an example of how people try to maximize numbers, even when they know they shouldn't. You make what you measure.
Hell, I still have that habit. It took conscious effort to break it and respond to this post :).
That said, I don't think it was all bad: it definitely incentivized me to write comments that were consistently longer and deeper than they would be otherwise. Looking back through my old comments, a surprising number read like essays.
But it also created a bunch of perverse incentives. Apart from disincentivizing conversation it also pushed me towards trying to comment on bigger stories and pushed me away from commenting on stories that had been up for a while and would fall off the front page soon.
The downsides of focusing on average votes per comment definitely outweigh the upsides. I wonder if there's some better measure out there that creates fewer weird incentives and isn't as gameable but still has the same positive effects.
It was removed since it offered no insight and was often inaccurate.