Weekly self-assessment
Computed from the archive and the request log every week, and published whatever it says.
128 requests from 4 distinct sources reached the METERED surface and none settled under the FULL conditions. Something found the paid routes and asked for the data; the payment rail is unexercised. The source list is unaudited - see the gate block for what was counted and what was excluded.
Demand
- Window
- →
- Collector liveness
green- Verdict corrections
- 3
- Records
- gap record · demand series
completeness 97.1% is at or above the 90% gate, AND 2 expected tick(s) never ran at all; 17 opened and never closed. A cycle that dies mid-run publishes nothing and writes no row: the endpoints it had not yet reached have no reading for that hour AND no filed absence either, so this count is the whole of that shortfall's trace; 2 ran out of wall-clock budget, leaving 56 endpoint-hour(s) unprobed and recorded per endpoint as Not probed / tick-budget-exhausted. The seed is too large for the cadence, or the archive got slower. Ours, not theirs. Green is a completeness reading against a threshold and is not a claim that the window was clean.
A published verdict is never re-cut. A correction is appended beside it.
2026-W37 · PARTIAL → INSTRUMENT-GAP
- What was published
- The collector liveness gate read GREEN at 97.1% completeness, so the week was graded on demand: 128 requests from 4 distinct sources reached the metered surface and none settled under the FULL conditions, which is the PARTIAL rung. The gate's own words named 17 cycles that opened and never closed and 2 that never ran at all, and read green over them.
- What the record holds
- The gate reads RED, and a red liveness gate makes the week INSTRUMENT-GAP rather than a rung graded on demand. Completeness has not moved: 97.1%, at or above the 90% threshold. What fails is the gate's second half — 19 cycles died or never ran in this window, 18 of them explained by 5 recorded incidents, and one, t20260901T134500Z, explained by nothing on any channel we hold. The week's demand series are reported below unchanged and are not evidence of demand either way, because they were measured by an instrument with an unexplained hole in it. This reading was taken 2026-09-08T12:16:25.573Z by the shipped gate (`livenessGate`) reading incident register version 1 over the same recorded gap population the published object was computed from. The week was not re-run and no series was recomputed.
Every week
The verdict that ran in 2026-W44 has been published; it covers 2026-10-19T02:07:18.878Z → 2026-10-26T02:07:18.878Z, the days before it ran — and it was delivered 0.0 hour(s) after its scheduled slot, so the days between that slot and the run that wrote it were graded by no scheduled delivery of their own. The week we are in, 2026-W44, is graded by the verdict that runs in the week after it, so the days since that window closed are not graded yet. Every ISO week from 2026-W34 to 2026-W44 has a verdict — 11 week(s), none missing.