Four signals tracked on every pipeline run: execution time (mean and standard deviation, by parallel and sequential mode), flakiness, test independence, and critical path coverage. Flakiness is measured, not guessed. A scheduled probe runs the suite 30 times in declared order and 30 randomised, scoring each test on pass-fail transitions. Independence is the ratio between the two pass rates - below 1.0 exposes coupling. Latest: 0.77% flakiness against a 2% target, independence 1.004, 100% priority-one coverage. Static site on GitLab Pages reading an append-only JSON history. No database.