Skip to main content

Unstable tests

An unstable test — a flaky one — passes and fails without the product changing. It costs more than a failing test, because a failing test tells you something and an unstable one teaches people to ignore results.

What this screen lists​

Everything a person has flagged by hand. It is deliberately not derived from run history: Sekee does not guess that a test is flaky because it alternated twice. Somebody looked, decided, and said so.

Flagging one​

From a test case, choose Mark as unstable. You are asked for:

  • A reason — what you saw, and why you think it is the test rather than the product
  • An assignee (optional) — who is going to look at it

The case then appears here until the flag is cleared.

Clearing one​

Open the case and clear the flag. Members can clear a flag they were allowed to raise — one of the few deletions a non-admin may perform, and it is deliberate.

:::note A flag re-raised keeps its original date Flagging a test that is already flagged is treated as editing the flag, not replacing it, so the date stays as the first mark. That is what you want when somebody adds detail to an existing flag. :::

Working through the list​

The list carries the standard toolbar, so you can filter by assignee, by folder, or by tag, and the filter goes into the URL to be shared.

A good rhythm is to treat the list as a queue with a ceiling: agree a number the team will not let it exceed, and fix or delete rather than let it grow. A flaky test nobody intends to fix should be deleted, not flagged forever.