fadaly.net/work/ciquality
SRE & INCIDENT

FLAKE.

18 CI pipelines instrumented with DORA + flake metrics. 6 pipelines flake more than 8% of runs. 2 have a deploy frequency below once per month.

A flaky CI is a stop sign you taught the team to run.

CI-011 · auth-service e2e 14% FLAKE
Avg run 18m · retries on every 7th run · top flake: race condition.
Quarantine flaky test, fix race, restore 30d clean window.