They changed patterns alot.
It started with basic hateful messages directed at her, or saying she should graduate, die, etc.
Once those patterns could be filtered by night bot, they started to add simileys to obfuscate the messages. At some point they started to just use articles from literature, wikipedia, news, etc. I think this was the most prominent, as it was really hard to catch automatically by just the message. It was also not the same message per bot, but it continued the text. At that point the goal was just to disrupt chat.
There were also other patterns that aimed to simulate a fight within chat with the spammers, so a lot of "members only chat please", "I hate the spammers", etc. To make users also say those things.
With the moderators and towards the end it returned to mostly hate messages, made harder to automatically detecht using icons. And with the youtube anti-spam measures, you don't see it.
On twitter it is a bit different. There they just include the hashtags in everything, from pro/anti-china news, pro american news, their own conversations, spam, hate, etc. basically just fill the hashtag with so much garbage that is is unusable.
They also did stuff like impersonate Cocos fans, etc. I would bet there is also a lot of stuff on the channels we cannot see, like DMs, the contact form, etc.