Yeah you could rely on that kind of metadata, but then people often don't use it in chat rooms. They're replying to a message in a chat room, but they don't click the "reply" button like they do here. Here's an example. As far as the metadata is concerned, these two messages are separated. However, they're clearly connected by the context. I'm not even sure a timer would consistently separate them correctly. People could return to a chat room hours, days after the fact and respond in this way and it would make just as much sense.