You receive a collection of sampled call stacks. Each sample contains a `threadId` and an ordered list of frames from root to leaf. Design an algorithm and data model for the following related queries. ### Part 1: Most Frequent Stack Return the distinct stack observed most often across all samples. If several stacks have the same count, return the deepest one. State a final deterministic tie-breaker when count and depth are both equal. #### What This Part Should Cover - A precise definition of stack equality - Canonical keys and frequency counting - Tie-breaking by count, depth, and a stable final rule - Time and space complexity ### Part 2: Per-Thread Results Return the winning stack independently for each thread using the same ordering rules. Explain whether identical stacks from different threads share storage or counts. #### What This Part Should Cover - Correct partitioning by thread - Reuse without accidentally mixing frequencies - Output behavior for threads with no valid frames ### Part 3: Scale and Representation Discuss how the solution changes when stacks are very deep, frame names are large strings, or samples arrive as a stream. #### What This Part Should Cover - Interning, tries, hashes, or other compact representations - Collision handling if hashes are used - Incremental winner maintenance ### What a Strong Answer Covers A strong answer begins by clarifying the sample representation because the original task description is terse. It gives a simple correct map-based solution before proposing compressed or streaming variants. ### Follow-up Questions - How would you merge partial aggregates from several machines? - How would you compare stacks if addresses must first be symbolized? - What if the winner is based on inclusive frame counts rather than exact-stack frequency?