# Practice 08 · Async outcomes

## A. Follow every state
Use checkpoint 08 to load normal, empty, malformed and wrong-shape fixtures. For each, record the visible status and whether cards appear. Retry a bad fixture, then recover by selecting normal data. Explain the difference between an empty successful dataset and a filtered zero-match result.

## B. Repair a race
Request A starts; the learner selects B; B resolves; A resolves later. A naive loader commits both results. Draw the incorrect timeline and the intended timeline. Add cleanup and an obsolete-work guard in a copy, or explain the supplied hook line by line. Show that an old rejection cannot replace the new success either.

## C. Transfer
Add a deterministic test where a first request resolves after a second request. Assert that the second result remains visible. Add an HTTP-error case separate from malformed JSON. Do not use a real API or arbitrary sleep to force timing. Explain what your mock proves and what still needs a real-browser check.

Completion evidence: four fixture outcomes, retry/recovery, race timeline and deterministic evidence. [Separate solutions](../solutions/08.html). Then attempt [assessment 02](../assessments/02.html).