Choosing capture windows with change control in mind
Useful latency captures often need elevated tracing or packet sampling. Change-control boards rightly push back. The compromise is not to measure only at noon — it is to define a narrow window that coincides with a known stress pattern and to pre-write rollback steps.
We bring a one-page plan: start and stop times, signals that abort the capture, and who sits on the bridge. Boards approve more readily when the path under study is named and the blast radius is limited to observer overhead.
Quiet-hour samples still help for baselines. They should never be the only evidence when the incident report says “every evening.”