Banning a centralized service like gmail is easier, but then you lose in productivity as 90% of your country's businesses are likely to rely on it.
As far as privacy goes, it's also easier to coerce admins of smaller nodes to disclose valuable information than fighting with a multinational foreign corporation.
This is not an argument against decentralization. It's just not immediately obvious that decentralization does not automatically lead to censorship resistance. To do that we need onion routing or mix networks as a base for all our communications, so that banning the network would be equivalent to disconnecting the ISP from the Internet altogether.
Not if such network was made purposefully to minimize transparency.
To say that it’s “much easier” to ban hundreds of IP addresses instead of a dozen IP addresses is ridiculous. It’s no easier to get IP addresses in a distributed system than it is to get them in a centralized system.
GMail is a funny example to point to, because email is arguably the most used digital communication method and it’s also decentralized. Try to fully shut down email in the US — where are you going to find a list of all ip addresses hosting SMTP and IMAP servers, and is that going to be easier than just blocking Twitter?
Shodan is a readily available database of such data. Just search for all hosts with IMAP, SMTP and POP3 ports exposed and then block those. Alternatively, a DIY scan of the whole IPv4 address range can be accomplished within hours or days.
IPv6 address space is harder to scan, but it's arguably safe to ban all such traffic since nearly nothing critical uses IPv6.
Edited to add: there is a tiny fraction of providers that would reject the emails from such a setup, but it's really not common and does not include large providers.
And email is practically the only decentralized system in widespread use these days. And it's not centralized, but close to it. You shut down 3 or 4 email services, 95% of its traffic is gone.
You don’t need to shut down the entire public email network (although it’s possible to do with DPI). You would only need to ban the servers that are involved in spreading of unwanted information.
If there is a decentralized service that is exclusively used by "perpetrators" then banning the service will have a greater impact even if the ban is harder to implement.