By contrast, if you were comparing my submitted password to a stored password, I could use your response time to know that I got the first n characters right. Since n is a prefix of the password, I've learned something about the password, I can lock down those characters, and then guess more, etc. That much I understand.
But the case you're talking about is a comparison of hashes. In that case, timing attacks can tell me that my guess's hash's first n characters match the password's hash's first n characters. So now what? How do I adapt future guesses to learn more characters? What have I learned about (a preimage of the hash of) the password?
If partial hash collisions were, in any way, helpful in guessing the full collision, Bitcoin mining algorithms would adaptively change their guesses as they discovered that their nonces only led to a partial match.
1. https://security.stackexchange.com/questions/111040/should-i...
It's telling that even the very smart person who did such a great write-up made a few mistakes again in initial attempts at demonstrating a corrected algorithm that actually blew the whole thing up again.
Because it's so subtle, the odds are very good that most current systems are fallible to this.
1. https://codahale.com/a-lesson-in-timing-attacks/