Integration test flakiness review

Integration tests in this repo run on real VMs against live ESS stacks, on hosts that vary in speed, distro, kernel, and concurrent load. The dominant failure mode is a test that passes locally and in the PR run, then flakes weeks later on a slower host, a different distro, or after an unrelated product change. This skill encodes the failure modes actually observed in a year of flake fixes (see references/flake-taxonomy.md for the evidence: every rule below cites real incidents).

Method