TRY FIRST — use fictional data and your own practice copy.
1 Map architecture and responsibilities; separate code,state,config,evidence.
2 Reject workers 0/True/8,string launch flags and unknown secret-like fields.
3 Trace v1->v2->failedv 3->v1. Test wrong hash,immutable version conflict and incompatible schema.
4 Reconcile snapshot 2/70000 versus current 3/75000. State RPO loss implication,not a promised RTO.
5 Calculate 18/20 success,nearest-rank p 95 and 95%error budget;reject duplicate job/extra token field.
6 Trace pre-commit rollback and rerun counts. Test conflictingA and 3 transient timeouts.
7 Recalculate invented egress 50 units versus 100 units. Review IAM/cleanup without real credentials.
8 Review optional Docker/CI artifacts; run local equivalent checks; mark unavailable runtime unexecuted.
9 Write a 250-400 word incident/evidence memo using the templates and capstone rubric.

Record:case | prediction | exact command | actual | mode | limitation | follow-up.
Mini-project extension:copy fixtures,add a new independent eventD/5000 and a test for 75000.
Preserve baseline controls. Repeat D and prove it does not add twice. Explain external-side-effect limits.
No live cloud resource,site upload,account change,production restore or alert sending is part of this lab.
