So you prefer

    arr.reduce((acc, el) => {
      if (el % 2 === 0) {
        acc.push(el * 2);
      }

      return acc;
    }, []);

over

    arr.filter(e => e % 2 === 0).map(e => e * 2)

The only advantage of the reduction as I see it is performance, but this is highly dubious and would need to be profiled for proof (I don't recall seeing removing a pass like this matter in practice). And if perf does matter, a for..of loop would be clearer and one-pass, not to mention async-compatible:

    const result = [];
    for (const el of arr) {
      if (el % 2 === 0) {
        result.push(el * 2);
      }
    }

Exercises like this illustrate why verbal technical job interviews are useful in the age of LLMs--a series of A/B taste preferences seems high signal and ripe for discussion: "Ah, so you're choosing reduce for perf... please describe a scenario you encountered where this made a measurable impact".