Ask HN: LeetCode vs. Scenario-Based Interview Questions – what's your take?

I'm curious to hear your thoughts on different styles of technical interview questions. Specifically:

    - LeetCode-style algorithmic puzzles and data structure problems
    - Scenario-based questions that simulate real-world situations (allowed to use the internet), like:
        -- Tracing a slow API response
        -- Identify possible causes of the intermittent failures
        -- Front-end Rendering Issue

Which do you think is more effective for assessing candidates? What are the pros and cons of each in your experience?

I'd love to hear from both interviewers and interviewees: - Which style do you prefer and why? - How well do these align with actual day-to-day work? - Any suggestions for improving the tech interview process overall?