DaniWeb moderators ban a member about once or twice a month for a variety of reasons. In most cases, it is because the member spammed his or her website more than once or twice, (reasonably common since we have such a large webmaster audience). If you read a bunch of other posts around here, you'll see a repeating pattern of me complaining about having such a hard time maintaining an advertisement-free webmaster community. It's that reason why there has been a ban on any site specific talk in these forums, for example.
We're probably out of the norm, though. I wouldn't expect non business or webmaster-related communities (basically, all consumer-oriented sites) to have such a problem.
my last banning experience came after a member insulted my moderator.
I have a music forum and I have some people post in there. As I always do, I require good meaningful posts, not just spammy and useless chat posts. So, I had a member that was a bit too "relaxed".
My mod PM-ed him and tried to explain that we appreciate his comittment, but would love seeing posts that are not one-liners and that don't go too much off-topic. Instead of just saying "OK, sorry" he replied with some words that made my mod freeze. He's a young man and he's used to hearing some "strong" language, but the guy was really "intense".
After seeing the content of that reply I banned the offender immediatelly.
Oh that could get annoying then :) I've noticed that a lot of times one of my members will create a new username if they've forgotten their password instead of simply requesting a new password to be sent to them. When that happens and I find out about it, I just merge the accounts so all the posts get combined and so forth. But creating multiple accounts to troll, or to create a second account because your first account was banned, is another matter entirely.
I don't like banning by IP because people connect via proxies and other things where multiple users share a single IP address. For example, many colleges do this, and it's a very bad idea to inadvertently ban the wrong people at a University - especially given the nature of these forums. And then there's AOL which is just a big giant proxy - I don't want to ban half of an ISP :)
Banning by IP is generally futile, anyways, because if a person is set out to troll, they will. They will just use a free proxy or anonymizer service on the web to continue posting under different IP addresses. So all that you're really ending up doing is banning the innocent guys.
Therefore ... We ban by username in nearly all cases. If we notice a troll keeps reregistering with multiple usernames, then we will ban by IP but only temporarily. After awhile, the troll will become bored with trying to get in (unless he's using a proxy, during which time more legal measures must be taken) and he'll give up. Then we'll unban the IP, just because it's never a good idea to keep an IP address permanently blocked since very few people in this world have true static IP addresses.