Performance smoke test

A performance smoke test is not a load test. It runs a few virtual users for a couple of minutes on each candidate build and compares response times for a named journey against a threshold. Its job is to notice that search became slow on this build, not to find the breaking point of the system.

The thresholds should come from what a user feels on that journey and from the last good build. A number with no journey attached does not help anyone decide.

I write these in k6 and keep them next to the API smoke checks. A full load test is a separate exercise with its own environment, data and owners.