# INFRA-REGISTER sweep prep — v0.69.0 / releases#294 close
Measured 2026-09-10 by doyle at main de5a44bc, during the golden 34445961595 wait window.
NOT yet acted on: the sweep trigger is RELEASE CLOSE, and golden's nine-row verdict is pending.

## Section boundaries (verified, not assumed)
`grep -nE '^## '` returns exactly three: 178 `## OPEN`, 1142 `## IN-FLIGHT ON THE BOARD`,
1148 `## BUILT / RETIRED`. So OPEN is the contiguous block 179-1141, one partition, no
intermediate heading re-partitioning it.

## Finding: 8 of the 20 OPEN entries are DISCHARGED by their own Status line
Population: 20 `### IR-` headings inside 179-1141 (IR-1..IR-14, IR-17, IR-16, IR-18, IR-19,
IR-20, IR-54 — note IR-16/IR-17 are out of numeric order in the file, and IR-15 already sits
under BUILT / RETIRED).

Discharged, belong under `## BUILT / RETIRED` (quoting each entry's own Status field):
- IR-2  RETIRED 2026-08-28 — measurement run, verdict NO FLIP, incremental stays ON.
- IR-8  BUILT 2026-08-18 on KEYSTONE #182 hygiene lane.
- IR-9  pin LANDED 2026-08-18 at 35968156, audited **BUILT** at the 2026-08-29 v0.65.0 close.
- IR-13 BUILT 2026-08-18 on KEYSTONE #182 hygiene lane.
- IR-16 BUILT 2026-08-03 (landed @62c5623, LOCKSMITH t1 lane).
- IR-18 BUILT 2026-08-04 — named acceptance carrier RAN, condition DISCHARGED.
- IR-19 RETIRED 2026-08-18 — intentional for main pushes, not a classifier defect.
- IR-54 BUILT 2026-08-29 — PR #176 (880b3b9c, landed on main via 37368263).

NOT discharged, stay in OPEN (residual named):
- IR-12 BUILT "(the narrowed blind arm only)" — the 2026-09-06 sweep header composes the
  residual as a W3 drift-gate rider held by hertz. Partial build is not a discharge.
Genuinely-open population after the move: 12.

## Why this is worth one commit, not a shrug
The mandate line is "doyle sweeps this file at every milestone intake and every release close."
Eight discharged entries inside the swept partition means every sweep since 2026-08-18 has
re-read ~40% of a population that had nothing left to decide — and a sweep that costs more
than it should is a sweep that gets done shallower. Relocation is the house form (the
BUILT / RETIRED section exists and already holds IR-15), so this is a MOVE, never a delete:
the entries' bodies are the measured record and stay verbatim.

## Known structural wart, do not "fix" blind
The 2026-09-06 sweep header records that IR-31 has NO `###` heading — its body sits under
IR-30's, flagged and deliberately not normalized. Any script that partitions the file by
`### IR-` headings will silently fold IR-31 into IR-30. That wart is in BUILT / RETIRED, not
in OPEN, so it does not touch this move — but a whole-file normalizer would hit it.

## Action at close (only after golden green)
1. Move the 8 entries verbatim to `## BUILT / RETIRED`, asserting byte identity of each moved
   block and of both untouched regions (the MEMORY.md split method).
2. Add the v0.69.0 sweep paragraph to the header's "Last sweep:" chain.
3. Then do the actual ripeness pass over the 12 remaining OPEN entries.