Systems like we are discussing can and do exist, I just don't see them as a necessary or particularly effective solution...especially when compared with the necessary trade-offs.
Look at AOL, which during it's hey-day was basically this sort of walled garden. You needed a phone #, name, address, payment combination to register. Many chatrooms were patrolled by moderators...there were even investigations and "stings" of trouble-makers. If you got TOS'd too many times you were permanently banned.
However trolling, illegal activity, pornography, etc thrived pretty well regardless. It was easy to evade the bans and there was a steady flow of new trolls.
I think the best solution in terms of respect for privacy and effectiveness is a combination of crowd-sourced moderation and intelligent detection of likely "trolling" and problem posters by software.