# Postmortem — The Incident Atlas

Preview only; no product adoption is implied.

## Current / change
The warm serif narrative and dark impact strip are coherent, but visually overlap the research family; the narrow phone timestamp gutter makes long events harder to read.

- Replaced the warm serif shell with humanist headings, grotesque prose and mono incident/time/data roles.
- Put the original impact facts beside the causal summary and broad, byte-identical source SVG.
- Moved timestamps above event text on phones; dark time plates distinguish the three source key events without inventing events.
- Retained both root-cause paragraphs, factors, successes and all action records; mobile keeps Owner/Due labels beside the original values.

## Benefit and trade-off
A compact engineering identity opens into impact accounting beside the original failure trace, then a time-indexed incident record with a complete root-cause panel.

Trade-off: Less literary than the current design. Full-width phone events and comfortable action records make the incident review longer.

Recommendation: prefer the new composition for this reader’s job; parent selects before integration. Keeping Current remains a valid choice.

## Content and labels
No new visible DOM text. At mobile, Owner and Due repeat for each of four actions (8 painted CSS labels). Other data-mobile-label attributes are defined but not painted. Original table headers and SVG accessible name remain intact.

Exact normalized source text-node multiset, numeric multiplicities, paragraph/heading/footer strings and keyed records pass; full source SVG bytes and meter widths remain unchanged. No factual content added, missing or changed. See evidence/content-parity.json for original records and counts.

## Visual review
The evidence spread has an engineering identity distinct from the approved serif Report. The complete curve and dashed annotations remain legible. At phone width, the time plate and event prose use separate full-width rows; the action table becomes an accountable record list.

The 768px image review caught a narrow Status column that wrapped In progress into fragments despite zero overflow. Reserved sufficient status width and prevented status-word fragmentation; rerendered all final evidence.

## Verification
Final captures at 1440/768/390/320, proposed boundary pairs, all existing/proposed boundary smokes, three labelled growth fixtures at 1440/390/320, forced font fallback at 320, computed text contrast and actual phone accessibility-tree checks. Body prose is at least 16px with line-height at least 1.5. Source and new contain no scripts, links, forms or disclosures; no no-op controls were added.

## Limitations
Chromium only; no Safari/Firefox or physical-device test. The long-title/thin/growth fixtures are labelled synthetic stress samples and are excluded from sample parity. No arbitrary document-height gate. Full-size screenshots and standalone HTML remain the reading references, not reduced comparison overviews.

Source SHA-256: `336ec7b129f1ade44e1b36c505f1efec847737651cbf6496cf29f3258ea665c2`
Preview SHA-256: `791e91c086761ac21825ceca2ccb13d86699209e9533ec91e179b96dd38815a9`
