Flaky test policy

Flaky suites are the fastest way to destroy trust in the gates — and once a team stops trusting the gate, the gate is worthless. At agent speed, that trust is load-bearing.

The policy

• Automatic detection — reruns and history identify non-deterministic tests.
• Quarantine lane — a flaky test is quarantined (still run, but not gating) with a fix-or-delete SLA — it doesn't sit quarantined forever.
• Agent-assisted triage — an agent proposes the cause (timing, order-dependence, shared state); a human confirms.
• Visible flake rate — a tracked metric (metrics); a rising flake rate is a problem to fix, not tolerate.

Why zero tolerance

A gate that fails randomly trains people to re-run until green — which means it's no longer checking anything. Protecting gate trust by aggressively quarantining and fixing flakes is what keeps the merge gate meaningful.