The latter is more important, I think: there is no reasonable way for a bad actor to set out to spam the REVIEWING of takedown requests. You could even make spamming takedowns more likely to trigger human review. So you can target a person on YouTube, you can try to do things to harass them, but it's not really feasible to try and attack the entire YouTube 'takedown review' process just to protect your attack on an individual.

That's already how it works.

How could this type of blackmail happen if someone was already reviewing the videos manually? Takedowns being automated is fundamentally at odds with free speech.

YT is legally required to accept all takedowns and takedown counter-notices to reduce liability. Not taking down something in lieu of a DMCA takedown request would put YT at risk of a lawsuit if the submitter does own said copyright.

Content ID, on the other hand, is a problem of their own making.

And how do you know that's not what Youtube is doing already ?

A sufficiently poorly done job is indistinguishable from nothing being done at all.

It's possible that it's impossible to do any better than they currently are, but as we don't have access to all of the data that they do, we cannot make that determination.