Having worked in this space for decades, I think most people I've met agree it's ludicrous, but there's strong pulls to keep it in place.

The most compelling argument I've heard is "we get so many applicants at $bigcompany that we need a fast and objective way to filter people out". And sure, that works, but think of the masses of great people you're turning off with that approach? Most great devs I know won't put up with that crap, including myself. Not a problem if you've optimized your company to build masses of code with early-career employees I suppose.

Knowing the tradeoffs we're making in data structures and having a broad understanding of different algorithms to throw at a problem is very handy. It's also almost completely unnecessary at the typical web shop (like you say). I use these things a bit in my work (not the typical web shop), but mostly indirectly via DBs and similar tools where I need to understand the tradeoffs. I'm certainly not implementing anything with red/black trees, making my custom bloom filters, etc. We have an ecosystem of tools for a reason, that would be silly to reinvent the wheel everywhere without a damn good reason.

One thing I realized after doing some studying is that your approach is limited by the concepts and tools you're familiar with. As I learned more about data structures and algorithms and as I practiced using them and evaluating the complexity of my implementations, I felt a lot more clarity and confidence about solving problems and structuring my code efficiently.