Yes, but my recollection is that the whale fail was usually due to huge growth or unprecedented traffic peaks. I don't think Twitter is growing at a fast rate any longer.
Rails was never meant to be run on such a massive load. It was the perfect framework to get twitter take over the market but once that goal was reached a rewrite was unavoidable with a tech stack that can sustain such a load.
"Record high" in an established business like Twitter isn't quite the same thing as a traffic spike in the growth phase of a startup (i.e. Twitter in 2011).
Record engagement these days is most likely measured in single digit percentage points over average engagement (at most maybe 10-20%), where when you're early and your traffic is spiking, it can be an order of magnitude more traffic than what you're used to seeing. The absolute numbers are larger yes, but from the perspective of "what can my infrastructure / architecture handle", relative changes are far more relevant.
> Musk claims often that Twitter engagement numbers are at record highs. I don't believe these are lies.
You're talking about a man who has more than once been fined by the SEC for making false statements about public securities. What makes you think that he'd be a more reliable narrator for a private company for which he has no external accountability and a large personal financial interest?
> At the very least per tweet it’s possible to see really view patterns, which to my knowledge what’s possible before.
Those numbers have been proven to be incorrect, by people who are posting from locked accounts with 0 followers racking up large nonzero view counts within a few minutes.
Some part of my brain immediately replied to this with the theory "what if those view counts are real and Twitter's just kind of ignoring whether an account is locked now" and damn that is a hell of a possibility to ponder.
There's another plausible theory that the just implemented it by counting database reads. So of course a locked test account will have more reads every time you look.
> There's another plausible theory that the just implemented it by counting database reads. So of course a locked test account will have more reads every time you look.
It'll have one, maybe two more reads every time you look. Not dozens. One person posted an example that had literally over a hundred alleged "views" within a minute, despite not refreshing their timeline after posting the tweet.