Validate per-item outcomes inside batch success

These examples and illustrative results are independently authored teaching materials, not measured model results.

Use case

A teaching batch requests 10 URLs and succeeds externally, with 8 complete, one DNS failure and one missing result. Match input identity rather than equating HTTP200 with ten successes.

Mechanism

Inspect actual envelope/item states, matching IDs/URLs. Preserve complete/failed/missing/pending and inspect duplicates/extras. Classify recovery within budget and report per-item scope without fabricated replacements.

Bad example

HTTP200 or ten objects means complete regardless of repeated URLs/errors/identity gaps.

Good example

Map all teaching inputs:8 complete, one DNS failure, one missing. Inspect inner status/content; outer success establishes transport only. Retain failed/missing identities and safe checks, never duplicated counts as coverage or invented text. Unexpected schemas remain parse gaps; reconcile identity and total.

Why the change matters

Batch envelopes hold mixed outcomes, and equal counts can hide duplication. Identity mapping accounts for every input while preserving usable/unfinished work.

Observable expectation

The fixture reports 8/1/1 matched to 10. Duplicating one success while omitting another fails identity despite equal total.

Pending is not complete and DNS errors contain no valid page; no real URLs are fetched.

Limits

Shapes/states vary; data.data is not universal. Item completion establishes neither content quality nor permission.