Quarantine

Detection on its own is just a report. Nijam tells you which tests are flaky, but until you act on that, the suite is still red and your team is still ignoring it.

The usual fix is test.skip() in the repo. That needs a pull request, it stops the test running, and nobody ever revisits it. Eighteen months later there is a pile of skipped tests and no one remembers which ones still matter.

Quarantine moves that decision into Nijam, where it stays visible and reversible.

What it does

A quarantined test still runs and still records results. What changes is only how its failures count:

• The run is not marked as failed when its only failures are quarantined.
• The run shows as a pass across your run history and pass-rate stats.
• The check Nijam posts to GitHub or GitLab stays green. If your team has that check set as required, that may unblock the pull request. Whether it does is your branch protection setting, not something Nijam decides.
• Slack notifications on the fail and regression triggers do not fire for a quarantined-only failure, because the build did not break.

A run with one real failure and one quarantined failure is still a failure. Quarantine subtracts from the blocking count, it does not silence the run.

It is never silent

Every surface that goes green because of a quarantine says so and names the tests:

• The pull request check and comment carry a Quarantined, not counted as a failure block.
• The GitLab commit status reads Passed, 2 quarantined tests failed.
• The run in Nijam renders as green-but-not-clean, never as a plain pass.
• Any member of the organization can see the full quarantine list, even though only admins can change it.

That is deliberate. A green check that quietly hides a broken test is worse than a red one.

Why it does not skip the test

Quarantine is a verdict, not a skip. Nijam keeps running the test on purpose, for two reasons.

A test that stops running can never prove it is healthy again. Because quarantined tests keep reporting, Nijam can tell you “this has passed 14 runs in a row, take it out of quarantine”. That prompt is the whole difference between a ledger and a graveyard, and a test.skip() in your repo can never give it to you.

It also means quarantine needs no reporter change and works identically for Playwright, pytest, and Vitest.

Using it

Quarantine a test from anywhere it appears: the run’s file view, the test explorer, the flaky list, the failing list, or the test detail page. Organization admins only, since it changes whether future runs are recorded as failures.

The full list lives under Test explorer, Quarantined, showing each test, who quarantined it and when, an optional reason, and how it has behaved on recent runs. Removing a test is a single click; its failures count normally again from the next run.

What it does not do

• It is not retroactive. Quarantining changes how subsequent runs are judged. Checks and comments already posted are never rewritten.
• It does not change the recorded result of the test itself. The test failed, the run detail says so, and the failure count still includes it.