Fwiw, I didn't practice for my tech interviews and found my Google new-grad interviews pretty easy. IIRC, the questions were "implement a cache" and "turn a sorted array into a binary tree". I haven't done work similar to algorithm problems for a decade now, and I'm pretty sure I could ace that interview again with my eyes closed, using skills that are relatively useful for a general SWE job. Though obviously I was much closer in time to the algo questions of my CS education, and most of the complaints I've heard about the process are not centered around new-grad interviews.
Though perhaps that's your point? As the size of the organization scaled, perhaps they moved away from the type of interview I got, towards interviews that more closely match your complaint. Conducting an interview that accurately predicts job performance is basically an impossible task, so I wouldn't be surprised if, speaking generally, the quality of their interviews suffered from their scaling just as much as their candidate pool did.
I wonder if their focus on mind-numbing algo questions in interviews also contributed to this? Somebody willing to spend months grinding Hackerrank, Leetcode or the like to get a job probably isn't a bastion of criticial thinking.