Answer: Data skew resulting from uneven distribution, where certain partitions contain significantly more records than others.

### Explanation The scenario described—where the **minimum** and **median** task durations are low and consistent, but the **maximum** duration is an extreme outlier (100x higher)—is a classic signature of **Data Skew**. * **Data Skew (Correct):** This occurs when a few partitions contain significantly more data than the rest. Since Spark processes partitions in parallel, the 'straggler' tasks handling these large partitions become a bottleneck for the entire stage. * **Task Queuing:** If thread pool configuration were the issue, you would likely see tasks waiting to start despite having available cores. It would not typically cause a single task to execute 100x longer than others once it has already started. * **Spillover:** While spilling to disk (due to memory pressure) does slow down tasks, a 100x duration difference is most commonly rooted in the sheer volume of data in a single partition (skew) rather than just the mechanism of writing to disk. * **Network Delays:** Latency from cross-region data transfers would generally increase the duration for all tasks performing reads, leading to an elevated median duration rather than a single extreme outlier.

Author: LeetQuiz Editorial Team