Before the production track
Pre-release Stability Gate
Promoting a candidate because the office devices did not crash is how last month’s Xiaomi cluster returns on Monday morning. The gate is a short, named reading of the testing track.
Before the production track
Promoting a candidate because the office devices did not crash is how last month’s Xiaomi cluster returns on Monday morning. The gate is a short, named reading of the testing track.
We do not run your test scripts. We read what the candidate already produces on the tracks you use: internal, closed, or open testing. If a signature from the last production version is still present, the note says hold and points at it.
Teams book a gate when a release includes an NDK bump, a new ads SDK, or a WebView-heavy screen. Those changes are where silent crash types tend to arrive.
Release owners who already have an internal candidate and want a second pair of eyes on vitals from internal, closed, or open testing tracks before promoting to production.
A go / hold note tied to named clusters, not a vague “looks fine.” Hold means a cluster still matches a known fatal or ANR signature, or a new native crash has appeared.
One release candidate (one version code or a short version-code range) on testing tracks. Production users are out of scope unless you expand to a full audit.
You tell us the version code, what changed, and which previous clusters you believe are closed.
We watch the testing track for an agreed window — usually three to five days of real testers, not a single smoke pass.
Go or hold, with signatures listed. Silence is not a go.
Testing-track access, mapping for the candidate, and the list of clusters closed since the last production version.
A gate with almost no testers cannot honestly say go. We will say the sample is too thin rather than invent calm.