How are you handling flaky tests in your CI/CD pipeline?

Flaky tests are killing our team's confidence in the test suite. We're seeing around 8–12% flake rate on our Playwright E2E suite in CI. The tests pass locally 100% of the time. Things we've already tried: - Adding explicit waits and asserting stable states - Running with 1 retry in CI - Isolating tests to run in separate workers The flakiness is concentrated around: - Intercepts that occasionally don't register fast enough - Network-dependent tests where the staging API is sometimes slow - Selectors that work against a live server but not against mocks What strategies or tools have worked for your team? Looking for systematic approaches, not just "add more waits".