Monkey tests
Each entry below is a dated run report. Open the report for the verdict block, method, findings, and artifact links. See QA reports for the reading and authoring conventions that apply to both streams.
Reports
Section titled “Reports”| # | Date | Scope | Build | Environment | Verdict |
|---|
No runs published yet. The first report lands here after the inaugural Twitch Ingress monkey run.
What we run
Section titled “What we run”- Lifecycle chaos. Kill, pause, and slow individual processes inside a service. Assert the supervisor recovers within its declared budget.
- Input fuzzing. Replay a recorded Twitch EventSub stream with bit-flipped, truncated, or out-of-order frames against a staging tenant.
- Load bursts. Short, high-rate event bursts to surface back-pressure and queue overflow handling.
What we do not run
Section titled “What we do not run”- Anything that touches a real broadcaster. Monkey runs use staging tenants with synthetic OAuth grants. A monkey run that hits a live broadcaster’s chat is a bug in the harness.
- Long soak tests. Soak runs live in the CI nightly job, not here. This subgroup is reserved for bounded, reproducible runs that produce a single artifact set.