Earlier quoted context omitted.
I don't think you should block a major domain because of a spam spike. If you think the entire domain is compromised, then send in a human to look, otherwise focus on the accounts that are sending spam links.
Meanwhile thousands of users are receiving links to malware and phishing attacks while human raters are twiddling their thumbs. The parent is correct, these systems should be automated. Automated systems can respond faster and more accurately than humans. Humans should only be involved to improve the system and correct any mistakes it makes after the fact.
If you want to prevent phishing links then you can't allow links in the first place. If you decide to allow links, then you need to add some stability to the system. Circumstantial reactions should not be fast. Banning a whole domain based on short term percentage is circumstantial.