The difference with IRC is that when one agrees to the global terms of the particular network they are joining, there are global network operators and admins that can ban bad behavior network wide. This is a middle ground between one centralized service like Twitter and the completely decentralized concept of Mastodon. There is still drama and bad actors but I would suggest far less than what one gets when each instance can have it's own rules or lack thereof. When a person is repeatedly violating the global rules they will be globally banned on that specific IRC network. They can go join another one and repeat the process.
Anyone can create their own IRC network and have their own global rules and connect their servers to their friends servers. There are small IRC networks, some are public and some are private.
Based on all the discussions I have seen on HN about Mastodon, I believe that the lack of a concept of global operators and admins is the missing piece that encourages bad actors to poison the well. The bad actors will find the instances that lack moderation and the human moderation delay will make the entire network look bad. This is not even counting the trolls that intentionally post illegal things and then instruct search engines, archive.org, archive.ph and other archive sites to mirror the content so they can get the domains and servers cancelled. To add more clarrification around my comment, this is not so much about the technology yes IRC lacks a web interface by default and must be bolted on but rather about the concept of global admins and moderators. People can block a specific instance but that is just and endless game of whack-a-mole and does not solve the root cause of the issues.