Salting isn't really supposed to make a hashing algorithm secure by being secret but by being unique. Unique salts make hashing more secure because an attacker can't re-use a single rainbow table for multiple hashed passwords. That, combined with a sufficiently computationally difficult hashing algorithm, it makes it prohibitively expensive to reverse the hashes of all your users.

This may not be enough to protect high value users or those who use fairly common or easily guessable passwords. This is part of why it is so important that you don't reuse passwords. It's also why your application should reject all known passwords using something like https://haveibeenpwned.com/Passwords or any of the "common password" list you can find online.

Edit: If you do include a secret that is stored seperatly that is added to the password and salt when hashing, this is called "peppering" and these peppers are generally not unique per user.

How long does it actually take in practice to break something like this? I would love it if someone could prove it to me.

With regards to crypto mines being used for breaking hashes, if you have one based on GPUs, yes, you could reuse GPU mining hardware for cracking hashes, albeit with relatively low hashrates for current best practice hashing algorithms.

If you're looking at something like Bitcoin's hashrate and thinking that it could be used to break SHA2 hashes, as far as I understand ASIC miners, this is not possible, as ASIC miners are designed only for mining, and they don't really accept non-mining related inputs (ie, no arbitrary inputs to be hashed, unless it matches Bitcoin's specific steps for iterating over nonces).

I'm really curious where people get their ideas about salting. It's not just a word. It doesn't make one password any more difficult to crack. It makes cracking every password in a given database more difficult to do. A password's salt is public information.