The early net code assumed trust. Email, DNS, FTP, etc. You were trusted or authorized to use particular computers. Getting on those meant talking to someone with a bit of sense not to let spammers/trolls onto their networks. Then if they did get someone like that they walked over to their terminal and just locked them out of their account. It was also why they heavily stayed away form corp interests for so long. As many times a small cohort of people will ignore doing the right thing if it gets them some sort of power/wealth. So we end up with central authorities who can either limit or kick off those who do trolly things.
It is not going back to that. It cant. But the trick is can we get those distributed properties back? Yet still keep the spammers at bay?
Most people who use these systems could not care one whit if it is centralized or not. They just want to send some cat pics and watch a video of someone doing something silly and give a thumbs up to that.
To me the best part of decentralized systems is that they allow for centralized sub-systems to emerge organically, instead of being decided top-down.
For instance, I don't think that when launching the first internet nodes we would have anticipated that certs should have centralized entities, it happened that it made more sense that way once we saw the issues emerging from actual use.