📝 docs(planning): record the 2026-08-25 operator rulings and withdraw the stale rule-7 block (D-162..D-164)

[konih]

Docs/spec only. Zero Go, zero schema bytes.

D-162 — P5-DEM is re-decomposed spec-first

Operator-ruled. All 15 DEM stories carry zero Verify: annotations, so not one
states how it would be proven. Treated as a spec defect (hard rule 4), not a per-lane
gap: DEM-S14 is the epic exit gate and has nothing to aggregate otherwise.

D-163 — the dependabot changelog recurrence is fixed at the bot; OQ-30's guard STAYS

I first asked the operator this question on an incomplete premise and they ruled to
enable the drift gate on pull_request. That ruling was withdrawn and re-taken once
the premise was corrected:

• The blocker is not RELSE-08. It is a deliberate if: github.event_name != 'pull_request'
  guard tracked as OQ-30, already ruled (b) keep the guard by D-148.
• Dependabot cannot run task changelog-write on its own branch, and verify is a
  required check with strict=true / enforce_admins=true (measured via the
  branch-protection API). Enabling the gate would convert "main red for ~1h, healed by one
  regen commit" into "every dependabot PR red on a required check until a human pushes a
  regen commit to the bot's branch" — strictly worse for the incident that prompted it.

D-163 also records, rather than leaving standing, that D-148's stated mechanism does not
survive: OQ-30 measured the merge-direction ordering hazard only in the lane-first-parent
direction, while GitHub mints the merge ref base-first. D-148 is upheld as precaution, not
as mechanism. The mechanism sub-choice is deliberately not taken and is parked in INBOX.

D-164 — DOC-TRUTH: the rule-7 block was withdrawn nine days late

D-144 (2026-08-16) resolved rule-7 to (d1) and ends "Unblocks E11-S03 and E11-S04."
Four documents still implied the question was parked on the operator, and that staleness
actively misled this session — the handoff and the operator's own opening message both
carried it forward as fact, and it was nearly put to the operator a second time.

Corrected: backlog.md E11-S03/S04, docs/planning/rego-tier-ceiling.md (written after
D-144 and stale on arrival), openspec/specs/p5-e11-rego-backend/spec.md REQ-E11-S04-03.
Retained analysis is kept; only the false implication is withdrawn.

Generalisation deliberately not made: no gate asserting "every red row has a live
blocker" — that predicate would match row TEXT, which is the defect species behind the
14 P1s of the last two waves.

INBOX

Retires the PR #85 block (merged e2e7691, D-161), records the rulings, and leaves exactly
one 🔴 item: the D-163 mechanism sub-choice.

Verification

task check EXIT=0 (all 21 stages); task changelog-verify clean after the commit exists.