Skip to content

docs(backlog): record the #1008 ruling, and #1234 as blocked by it - #785

Merged
wshallwshall merged 6 commits into
mainfrom
claude/contractor-533440
Sep 4, 2026
Merged

docs(backlog): record the #1008 ruling, and #1234 as blocked by it#785
wshallwshall merged 6 commits into
mainfrom
claude/contractor-533440

Conversation

@wshallwshall

@wshallwshall wshallwshall commented Sep 3, 2026

Copy link
Copy Markdown
Collaborator

What this is

BACKLOG #1008 has spent three passes asking for an owner ruling on the substance and never had one. This records the ruling. Nothing else changes: the gate stands, the banner does not flip, no code is touched.

Why the record needed this specifically

The item is unusually clear about what would and would not lift it. Its 2026-08-12 entry records the owner, asked to choose, answering "do what you judge best" and states plainly that this is a delegation rather than a ruling, and must not later be cited as "the owner decided the gate stands". It also records the prerequisite-fired argument being offered and rejected twice, by different lanes, and warns that a session reasoning from 1e9cc4c1 to a lift has re-derived a rejected argument.

So the gap was never evidence. It was an actual decision. That now exists.

What was put, and what came back

Put by the Contractor seat on 2026-09-03 with the three live options stated:

  • keep the preflight deferred
  • lift the gate and build it
  • fix only the ungated probe defect

The owner chose keep it deferred. Grounds as put: the runbook fix removed the cause and has landed, the preflight detects a symptom whose cause is gone, and at zero deployments demand for the detector is weak.

What the entry is careful NOT to decide

Two questions would otherwise be read as settled by this, and neither is:

  • #1234, the SQL Server probe returning OBSERVED unconditionally while NULL role results fold to False. That is a clean bill of health from a probe that read nothing. It is ungated, and it stays worth fixing whichever way the gate ever goes.
  • Blocker B, the delivered refusal being narrower than the item's stated Scope. That is moot while the gate stands, and it returns if the gate is ever lifted. Moot is not resolved.

Checks

  • scripts/docs/backlog_status_check.py passes: 657 items, each declaring exactly one status.
  • No new status glyph introduced. The added lines are blockquote prose inside the existing banner block.
  • Pre-commit ran clean on commit, secret detection included.
  • No Python changed, so mypy and the test suite are not implicated. Diff is 16 added lines in docs/BACKLOG.md.

I applied the reviewed label to my own work. The owner has asked not to use the Reviewer role, so that label records that a step happened rather than an independent read of the diff.

🤖 Generated with Claude Code

wshallwshall @claude
docs(backlog): record the owner ruling on #1008 -- the preflight stay…
3d15187 
…s deferred

#1008 has asked three times for a ruling ON THE SUBSTANCE and never had one. The
2026-08-12 entry records the owner answering "do what you judge best", which the
coordinator correctly refused to read as a decision; the item says in terms that
nothing short of a substantive ruling should flip the banner, and that a session
reasoning from the cleared prerequisite to a lift has re-derived a rejected argument.

Put to the owner 2026-09-03 by the Contractor seat with the three live options
stated -- keep it deferred, lift the gate and build the preflight, or fix only the
ungated probe defect. The owner chose keep it deferred. The gate stands and the
banner does not flip; what changes is that the ruling now exists on the record
instead of being inferred.

The entry also states what the ruling does NOT decide, so neither question is read
as settled by it: the SQL Server probe's false-clean defect (#1234) is ungated and
stays worth fixing, and Blocker B is moot while the gate stands rather than resolved.

No status glyph is introduced. backlog_status_check.py passes at 657 items, each
declaring exactly one status.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@wshallwshall wshallwshall added the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 3, 2026
wshallwshall @claude
docs(backlog): record #1234 as blocked, and the coupling that blocks it
8313edf 
#1234 was offered for dispatch. It is unstartable, and I re-measured rather than
trusting the 2026-08-23 amendment that says so, because that amendment is three
weeks old and this seat spent the day finding records that had gone stale.

It has not gone stale. It has got worse. require_least_privilege returns 0 hits
across the engine Python on main, against a positive control of 2 files for
require_managed_identity on the same instrument; store/privilege.py is absent;
and w3-store-privilege-preflight is 4 ahead and 532 behind with no pull request,
against 328 behind when the amendment was written.

The part neither item states is the coupling. #1234 is right that a code defect
must not be hostage to a policy decision, and on policy it is not. On startability
the two are joined anyway: the defect exists only on that branch, the branch lands
only if the preflight lands, and the owner ruled today that the preflight stays
deferred. So the defer leaves #1234 with nowhere to land. Both clearing paths the
amendment names reverse that defer, which is why neither is a builder's call.

Recorded on both sides so the next seat reads it from either end.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@wshallwshall wshallwshall changed the title docs(backlog): record the owner ruling on #1008 -- the preflight stays deferred docs(backlog): record the #1008 ruling, and #1234 as blocked by it Sep 3, 2026
@wshallwshall wshallwshall added reviewed A reviewer has read this. Removed automatically when new commits arrive. and removed reviewed A reviewer has read this. Removed automatically when new commits arrive. labels Sep 3, 2026
wshallwshall @claude
docs(backlog): file #1430 -- the in-hold sampler yields one sample, s…
9e68d35 
…o the window has zero width

Surfaced by the #1420 measurement, which could not have found it by reading. The
in-hold sampler produces exactly two samples per cell, one in-hold and one
post-drain, in 20 of 20 cells at hold 1.5 and 3.0. So the window five sites
describe as "first to last in-hold samples" does not span a slightly-wrong range.
It spans no range at all.

Filed separately from #1420 rather than as an amendment, because it is a different
defect with a different fix and it is the reason #1420's fix (a) cannot be built
yet. Excluding the final sample trips the len(samples) < 2 guard, so the excluded
arm is undefined rather than smaller.

The part worth reading twice is that fix (a) fails silently: run end to end it
zeroes wall #3 while the live SLO still reports ok=true, which is the ADR 0158
shape arriving inside the tool the project uses to detect that shape. Hence the
build order recorded here: guarantee two in-hold samples first, then narrow.

The cause is explicitly unmeasured. The reload probe is ruled out at 0.04 to 0.20
seconds; the FD process-table walk is a hypothesis with nothing behind it, and the
item says so rather than letting a candidate harden into a finding.

Also records that #1420's 65.5s is a worst-case bound and the observed tail is 1.08
to 1.24s, so the bound is not re-quoted later as a measurement.

Number allocated with alloc.ps1. backlog_status_check passes at 658 items;
backlog_citation_check passes; the ADR 0158 link target verified on disk.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@github-actions github-actions Bot removed the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 3, 2026
@wshallwshall wshallwshall added the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 3, 2026
@github-actions github-actions Bot removed the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 3, 2026
@wshallwshall

wshallwshall commented Sep 4, 2026

Copy link
Copy Markdown
Collaborator Author

LANDER: applying reviewed under an explicit owner override. The owner has suspended the reviewer process and directed that this label be applied as needed to clear the gate.

SO READ THIS LABEL AS A BYPASS TOKEN, NOT A REVIEW. It records that the gate was cleared by instruction. It does not assert that anyone read this diff, and no later reader should infer one from it.

What I can say: this PR appeared in a read-only pre-verification sweep of all 59 non-draft open PRs (per-PR defect read, plus pairwise git merge-tree against every PR sharing a non-BACKLOG file), and no blocking defect was recorded against it. That sweep did not run any tests, and CI has been down for hours, so nothing here has a current green.

-- Lander

@wshallwshall wshallwshall added the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 4, 2026
wshallwshall
Merge origin/main, keeping both new backlog items
eb0e6db 
Both sides append a new section at the same point in docs/BACKLOG.md, under different
item numbers. Neither supersedes the other, so taking a side would drop a filed item
outright. Both blocks are kept, in the ascending item order the file already uses.

Verified with parse_items from scripts/docs/backlog_status_check.py rather than a
hand-rolled scan, and by an ADDED/LOST item-set difference against both parents, which
is empty. A count alone would not catch two blocks folding into one.
@github-actions github-actions Bot removed the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 4, 2026
@wshallwshall wshallwshall added reviewed A reviewer has read this. Removed automatically when new commits arrive. and removed reviewed A reviewer has read this. Removed automatically when new commits arrive. labels Sep 4, 2026
wshallwshall
Merge origin/main, keeping both new backlog items
320ca29 
Both sides append a new section at the same point in docs/BACKLOG.md, under different
item numbers. Neither supersedes the other, so taking a side would drop a filed item
outright. Both blocks are kept, in the ascending item order the file already uses.

Verified with parse_items from scripts/docs/backlog_status_check.py rather than a
hand-rolled scan, and by an ADDED/LOST item-set difference against both parents, which
is empty. A count alone would not catch two blocks folding into one.
@github-actions github-actions Bot removed the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 4, 2026
@wshallwshall wshallwshall added the reviewed A reviewer has read this. Removed automatically when new commits arrive. label Sep 4, 2026
@wshallwshall
wshallwshall added this pull request to the merge queue Sep 4, 2026
Merged via the queue into main with commit 20ee0fd Sep 4, 2026
41 of 42 checks passed
@wshallwshall
wshallwshall deleted the claude/contractor-533440 branch September 4, 2026 19:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

reviewed A reviewer has read this. Removed automatically when new commits arrive.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@wshallwshall