the-custodian/docs/recovery/cache-only-disposition-2026-08-28.md
codex 6cd450fa55
All checks were successful
CI Smoke / host-smoke (push) Successful in 1s
CI Smoke / container-smoke (push) Successful in 2s
finish CUST-WP-0068: recover, disposition, discard the cache
Close T02–T06 from live hub evidence, disposition clay-borg product
records as not workplans, lift the task-reset restriction in state-hub,
dump and remove the retired local Postgres instance, and record ADR-010's
work-record divergence as closed.
2026-08-28 21:06:25 +02:00

2.9 KiB
Raw Blame History

Cache-only work-record disposition — 2026-08-28

Coordination: CUST-WP-0068. Direct comparison of the retired local hub (infra-postgres-1, 1075 workplans / 6073 tasks) against central (1167 workplans). Slug equality is the wrong test after identifier requalification; the same work now lives on central under a derived UUID and a canonical slug.

T02 — bootstrap-eligible repositories

All six named repositories now have file-backed workplans on central:

Repository Files (top-level) Hub workplans
rapp-openbao 2 2
policy-nexus 3 3
whitehat-security 5 5
ops-mason 5 5
ground-game 10 9 (+ ad-hoc bound as ground-wp-adhoc-2026-08-08)
zone-engine 2 2

The original blockers (missing task UUIDs, partial projection, dirty worktrees) are gone. No owner action remains.

T03 — non-conforming identifiers

Renamed at source and present on central:

Old id (cache slug) Live id Hub slug
GLAS-0001 GLAS-WP-0001 glas-wp-0001
REIN-A-00010004 REINAH-WP-00010004 reinah-wp-00010004
REIN-OW-0001 REINOW-WP-0001 reinow-wp-0001

RAIL-BS-ADHOC-2026-07-27 has no live file. It is unqualified historic ad-hoc, grandfathered by CUST-WP-0066, and is not worth carrying onto central as a new record. Cache copy retained only until the cache is discarded.

T04 — clay-borg

Fifty files under clay-borg/workplans/ use kind: product|meta|mixed and status: done. They are not workplans. Fleet tooling correctly ignores them (type: workplan is required). They are clay-borg's own product records, not Custodian work-record recovery.

Disposition: do not rewrite, do not project as workplans. The files remain the owner's source. The 50 cb-wp-* cache rows are not unique fleet work.

T05 / T06

Central held 669 workplans when this workplan opened and holds 1167 now. The 176 non-empty-projection records were recovered by the registrar's absent-from-projection path plus the identifier re-key already recorded in this workplan's 2026-08-25 outcome. T06's ruling was that re-key, not a second decision. Remainder is T03/T04 above.

Cache-only residue (not load-bearing)

Of 190 cache slugs absent from central:

  • 50 cb-wp-* — T04, not workplans
  • ~17 railiance-wp-* — prefix migration, retired on central (T09)
  • requalified ad-hocs and renamed identifiers — live under new slugs
  • title slugs and pre-ADR-007 labels of records recovered by UUID re-key

No open work record exists only in the cache.

T09

reset_repository_projection now syncs identified tasks of existing workplans. Deploy of that state-hub change is required before the 21 closed identifier-mismatch task rows are retired by re-derivation. They are all already closed (0 open), so they are not a live surface. record_id-less rows stay out of this pass, as scoped.