> For example, one situation where you want an intentionally slow algorithm is when dealing with passwords.

This is true, but there's also more to the story. Modern password hashes like Argon2 force attackers into a "time-memory tradeoff", to try to reduce the advantage that specialized hardware has over the general-purpose computers that human beings use. I find that a lot of folks have memorized a summary like "slow hashes are good", and that's doubly unfortunate, because 1) like the article says, fast hashes are good, and 2) slow hashes in and of themselves are no longer the best we can do for password security. I often wish that password hashes weren't even called "hashes", because it's just a very different problem that they're solving.

> or even just a high number of rounds of SHA-512

Please god no :)