The first interview was a great experience overall. I asked clarifying questions, and the interviewer always gave me exact answers. Following his instructions, I completed two versions of the Rate Limiter. The second interview was disastrous. The interviewer could barely speak a complete sentence when I asked him to clarify things. I double-checked and triple-checked with him, and all questions were answered with a simple 'yes' or 'no'. The question was trivial: ranking letters by multiple votes, a simple sorting algorithm. I asked for a tie-breaking strategy; he said there was no need to consider that. I then asked how to calculate the points if that started with the vote arrays’ length in descending order. He said yes. Yet, after I completed the code, he said no, the score should always start at 3, regardless of the vote array's length, and made me hardcode the magic number 3. This is a code smell, isn't it? He then said, 'Let's now consider breaking ties,' and gave me two strategies. However, these two strategies only described the first place. I asked what about the rest of the letters. He said I didn't need to consider that case. I did what he requested, and he asked why I was outputting an array. Sorting an array by ranks is supposed to return an array, isn't it? He then said, 'No, I only want the first place.' After I completed the code as requested again, he asked if I could optimize it. I explained that if he only needed the first place, we didn't need to sort. We could use one variable to keep track of the first place, achieving O(n) complexity compared to sorting's O(n log n). He said, 'No, I want you to sort it.' Finally, he asked me to make a code change from letters.sort(comparatorFn) to Object.entries(ranks).sort(comparatorFn).map(([letter, rank]) => letter), which made no sense at all. It was neither efficient nor clean.

Implement a Rate Limiter, sliding window, and token bucket.

Sort letters by multiple votes.