Earlier quoted context omitted.
They are literally throwing away revenue with every denied impression. An act of desperation obviously. My take: they broke something and need to shed load to keep the site running. The "extreme scraping" thing is the usual Musk BS.
I find the scraping explanation plausible. Some search engine bots are aggressive. With all the AI hype, I first thought of Microsoft Bing scraping Twitter at full datacenter speed to suck in more information for OpenAI.
But given that they now require users to be logged in, it should be computationally cheap to drop unauthenticated requests at the front door before they incur real expense.
It'd also be cheap to just blackhole datacentre IP space.
The sort of attack that would require this level of limits is malware on tens of thousands of residential machines that can use a user's existing Twitter session cookies. I'm really skeptical that's the case.