BYTE NEWS — integrity report for the `calendar` dataset
========================================================================

No recorded gaps in the last 14 days.

SCOPE
  Every dated event this site knows about, in all four provenance classes, past ones included. It is NOT a complete unlock calendar: contract-read unlocks are an allowlist of contracts whose shape is standard, and no third-party unlock list is imported. The `observed_*` columns are filled only for rows whose address this site watches.

COVERAGE — when this dataset's producers were not looking
  ledgers read: onchain  (0 ledger row(s), merged into 0 window(s))
  (none recorded in this window)

DENSITY
  Not a continuous record. A day with no rows on this dataset is a quiet day in the world rather than a gap in ours, so no per-day expectation is stated.

PROVENANCE — what produced these rows
  US Bureau of Labor Statistics — CPI release schedule: 13 row(s)  — https://www.bls.gov/schedule/news_release/cpi.htm
  Federal Reserve — FOMC calendar: 56 row(s)  — https://www.federalreserve.gov/monetarypolicy/fomccalendars.htm
  Typed by the desk (config.yaml): 0 row(s)  — Every curated row carries a source link; one without is dropped rather than rendered.
  Contract reads (allowlist): 0 row(s)  — Vesting contracts of a known standard shape, read on chain. A reading nothing has confirmed inside the staleness window is not drawn at all.

CORRECTIONS — what has been revised since publication
  (none)

OUR OWN DOWNTIME — recorded incidents inside this window
  2026-08-30      33 min  degraded  The wire wrote no ledger row for 33 minutes (stage 'budget')  [NOT YET WRITTEN UP]
  2026-08-28      43 min  degraded  The wire wrote no ledger row for 43 minutes (stage 'budget')  [NOT YET WRITTEN UP]
  2026-08-28   2,776 min  degraded  The price tape's dense writer wrote nothing for 46 hours (daemon CWD unlinked by a re-clone)
  2026-08-26      78 min  degraded  The wire wrote no ledger row for 78 minutes (stage 'budget')  [NOT YET WRITTEN UP]
  2026-08-19   1,264 min  down  The wire published nothing for 21 hours — a shell variable used before it was set

WHAT THE CHECKS DID NOT CHECK
  gate of 2026-08-31: 2705/2710 checks passed, 4 skipped, 5 failing, 0 block(s) cut short
    SKIPPED: digest-2026-08-09-pm: its links resolve to published posts and to pages on disk (3 checks) — published before the wrap became a linked recap — the stored post is the audit trail and is never rewritten
    SKIPPED: the display rung (44) is asserted on the lead card's figure — the figure is the hero number INSIDE the card graphic — SVG user units in an 800x450 viewBox — so its rendered size is `units x (card width / 800)` and no one value makes it 44px on the front page, an article page and a share kit at once. Measured: 126px before this batch and 68.7px after, at 1440x900, because §6 put the lead in a 1/3 column. Reaching 44 means re-composing the card graphic, which is a batch and not a line — see docs/BATCH-K-REPORT.md
    FAILING: …and the tape carries PLAIN-TEXT rows as well as linked ones — a row with nothing behind it has no `href` at all, so the two states are distinguishable in the markup and not only in the styling (batch L §9), and a page where every row linked would not be exercising that rule
    FAILING: every recorded incident carries a written post-mortem
    FAILING: …and the timeline shows we were first
    FAILING: …which keeps the permalink, grows the timeline, and records the event that ended it
    FAILING: …and the measured row is 46px on a phone and 42px above 620 — the heights §1 asked for, read off the rendered page

This report is assembled from records this site already keeps; it adds no new collection.
The datasets it describes: /api/v1/calendar.json and .csv