Everyone sizes the pool on average run length, and average is the one statistic an agent fleet does not have. Simulated, the p99 sat at 160 seconds whether the system was half loaded or nearly saturated, which means capacity was never the problem.
Updated Aug 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.
Everyone sizes the pool on average run length, and average is the one statistic an agent fleet does not have. Simulated, the p99 sat at 160 seconds whether the system was half loaded or nearly saturated, which means capacity was never the problem.
Almost every candidate reaches for capacity, and capacity is usually the wrong lever. The measurement that decides this round is whether the p99 responds to load. If it does, you have a queue and more workers help. If it does not, the tail is the workload itself, and adding workers moves the median while the number the customer is complaining about sits exactly where it was. Listen for whether they check that before proposing anything.
No comments yet — be the first to share your approach.
