Conversation
…tion links The census pinned its streak head to Runfusion#3158 / 1762 red. Re-measured 2026-09-29: Runfusion#3162 is the last completed red run and Runfusion#3163 is in progress, so the completed streak is 1766 (total_count 1767), with Runfusion#1396 (2026-07-26T04:14:37Z) still the last green and zero successes since. Also corrects three claims that measurement falsified: - The "frozen at Runfusion#3158" reading was an artifact of reading the head of a >1700-run result set off a per_page=100 page. Runs Runfusion#3159-Runfusion#3163 have landed since, all failures, so the lane is still exercised. Replaced with the trap that produced it, plus a note that the list endpoint can return an inconsistent snapshot across per_page values. - The method line described only the engine shards' FAIL shape. Verified against run Runfusion#3162's logs: the project column is a vitest project name present only in shards 1/2, the CLI shard puts a pnpm stream prefix before FAIL, and core has no column at all. A project-column-only parser reads 217 of 233 cases and drops shard 4/4 without erroring. Also recorded that bare substring FAIL matches ACTION_FAILED, SCREENSHOT_FAILED and ERR_PNPM_RECURSIVE_RUN_FIRST_FAIL, and that the logs need grep -a. - "Two open links" are both closed (the port landed; main now produces runs), so the operator-visible surface exists and is exercised while the lane stays red. A run existing is not a run turning green; the green-run confirmation is still outstanding. Census inventory is unchanged and re-verified: 234 FAIL lines / 233 named cases / 117 files, 233 per-case rows intact.
The headline asserted 1766 consecutive red runs in the present tense, so the figure read as a property of the lane rather than a reading taken at a moment. It is not: every main push extends the streak, and the count was already 1770 at 2026-09-29T08:13Z when re-measured with the doc's own command. A reader who copies the number forward states a falsehood -- the doc's own three-traps block warns about exactly this class of stale measurement. Re-measured head: Runfusion#3166 (a6e65a5, 2026-09-29T07:29:40Z) is the last completed red run, Runfusion#3167 was in_progress, status=success is still 0 across the window. Restate 1770 as a dated measurement, require the as-of run number and timestamp whenever the streak is quoted, and name the non-drifting part of the claim (zero successes since Runfusion#1396) as the durable fact. Census untouched: 234 FAIL lines / 233 named cases / 117 files.
ThreatCrush Security Scan4589 finding(s) HIGH/CRITICAL: 43 | MEDIUM: 4035 | LOW: 511
…and 4539 more. Full results in the Security tab. Snippets are redacted; ThreatCrush never prints matched credential material. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Docs-only correction of the FUSI-020 main Full Suite census.
What this changes
One file:
docs/solutions/test-failures/main-full-suite-census-2026-09-25.md.The census pinned its streak head to Runfusion#3158 / 1762 red runs and asserted, in the
present tense, that the number was stable. Measurement falsified three claims:
The streak head moved. Re-measured 2026-09-29: the completed streak is
1770 red (
total_count1771 with one in flight),status=successis0 across the window, and [FN-5978] Fix: Goal retrieval tools for agent citation Runfusion/Fusion#1396 (2026-07-26T04:14:37Z) is still the
last green. The "frozen at fix: the last 4 literal move targets — Retry threw on renamed boards; allow-list now empty (#3150 closed) Runfusion/Fusion#3158" reading was an artifact of reading the head
of a >1700-run result set off a
per_page=100page — runs fix(core): a type that taught the wrong invariant — staleness signal column narrowed to legacy ids Runfusion/Fusion#3159–gate: enforce the quarantine deletion ratchet — nothing ran it, and it could not fail Runfusion/Fusion#3167 havelanded since, all failures, so the lane is still being exercised.
The headline is now a dated measurement, not a property of the lane.
Every main push extends the streak by one, so any figure copied forward is
false the next day. The count is stated as 1770 as measured
2026-09-29T08:13Z, with the as-of run number and timestamp required whenever
the streak is quoted, and the non-drifting fact (zero successes since
[FN-5978] Fix: Goal retrieval tools for agent citation Runfusion/Fusion#1396) named as the durable claim.
The method line described only the engine shards' FAIL shape. Verified
against run fix(core): an archived child kept blocking its parent's delete on a renamed board Runfusion/Fusion#3162's logs: the project column is a vitest project name present
only in shards 1/2, the CLI shard puts a
pnpmstream prefix beforeFAIL,and core has no column at all. A project-column-only parser reads 217 of
233 cases and drops shard 4/4 without erroring. Also recorded: bare
substring
FAILmatchesACTION_FAILED,SCREENSHOT_FAILEDandERR_PNPM_RECURSIVE_RUN_FIRST_FAIL, and the logs needgrep -a.What is deliberately NOT claimed
Both "open confirmation" links are now closed — the port landed and main
produces runs. But a run existing is not a run turning green. The
operator-visible confirmation required by the card is still outstanding: no main
push Full Suite run has turned green since Runfusion#1396 on 2026-07-26. This PR does not
assert that confirmation.
The census inventory is unchanged and re-verified: 234
FAILlines / 233named cases / 117 files, all 233 per-case rows intact. This change corrects the
streak measurement and the method description, not a single disposition.
Diff shape
@@ -34,37 +34,92 @@— the streak/method rewrite.@@ -281,14 +336,16 @@— confirmation-link and follow-up-family text.Both hunks are in the document body. No code, no config, no test files.
Relationship to open PR #29 — read before merging both
PR #29 (
docs/fusi-020-census-ledger-frame,bd9e25ee6) is complementary, notsuperseded, and this PR does not replace it. The two are siblings off the
same base
d3573cca7and their hunks are disjoint:PR #29 pins the quarantine-ledger claim to a
head_shaframe; this PR pins thestreak count to a dated as-of measurement. They address different sentences and
can merge in either order. They are not a red→green replacement pair.
Merging this PR first is safe: the reviewer's own finding was that both CLEARED
items and the as-of defect are fixed, verified first-hand at these exact refs.
Verification
tree, and commit SHAs all match the locally reviewed values
(
5cd8a4396→0e52f1687), with author/committer identity and timestampspreserved.
d3573cca7..0e52f1687: +87 / −30, one file.20/9then75/29, one file each.234/233/117, and the second totals row233/117/17) arebyte-identical to
main.Reviewer
Cleared by Workflow Reviewer (
agent-9eac734a) at these refs, verifiedfirst-hand rather than from the author's report.
Refs:
0e52f1687, parent5cd8a4396, 2 commits offd3573cca7.