A billing 403 became a Bug, two Stories and a clone across two Jira spaces, all with the same description pasted in — and the test data landed on the one ticket that never reached QA.
Bundle availability was channel-driven, so I looked for a channel selector in the cart. There isn't one, by design — and the same mistake decides what belongs in a regression scope.
A sandbox with zero Role-to-queue mappings against nine in production, and a fix pushed to prod whose defective channel could only be exercised after release. Two ways a green result means nothing.
The expected results for a discount-gating story were already enumerated in a sibling story under the same epic — including a struck-out line that handed over an entire negative case.
A 40% discount calculated on the catalogue price instead of the promo price dropped a cart to £0. The bug was cancelled, and the reason was sitting in its own Expected Result field.
A requirement running a week late was ignored for three sprints and blocked the whole stream in the fourth. Why small risks never get raised, and what happens to the tester who raises them anyway.
New campaign bundles arrive every few sprints. The four test archetypes never change — only the promotion table does. That is a generator, not an authoring task.
A single Hypercare defect on a shared object needed fixes in three separate Jira spaces, and the only artefact tying them together was the originating Story. Then the same failure came back scored as Lowest.
After an amendment the original Order goes Superseded, a new Order carries the truth, and a fresh Orchestration Plan moves with it. Testing the old record is a silent false result.
Three defects, one mistake: the assertion landed on a state that was real but not final — an async close, a two-stage closure, and an API 200 followed by an auto-cancel.