RESERVE BLOCK RECONCILIATION

C-397 — Reserve Block collision reconciliation witness¶

Cycle: C-397
Status: EVIDENCE RECOVERED · TRACK R REPAIR NOT APPLIED
Source audit: Terminal Actions run 29710940106
Source SHA: 346cdc13630fa9415cef9d95ede95dcfdb59c721
Audited: 2026-07-20T01:28:54.815Z

Finding¶

The Reserve Block incident is a lineage collision, not corruption of the preserved seal bodies.

Do not read 194 as canonical_reserve_blocks. It is the verified C-377/C-397 reconciliation candidate count from the examined attested corpus only. Constitutional canon remains canonical_reserve_blocks = unresolved until Track R completes (see below).

The 125 pair count is explained exactly:

• 121 contested positions have two candidate seals → 121 collision pairs.
• Blocks 1 and 2 each have three candidates → 2 collision pairs per position.
• 121 + 2 + 2 = 125.

Contested ranges: 1–33 and 42–131.
Clean ranges: 34–41 and 132–194.

The complete 125 pair witness (kept/dropped seal IDs, cycles, quorum, timestamps, divergence flag) is preserved in C397_RESERVE_BLOCK_COLLISION_WITNESS.json.

Index cardinality vs examination vs constitutional canon (live, 2026-08-08)¶

Production vault status exposes three different lenses — confusing them caused this incident:

360 indexed seal IDs          ← vault:seals:index:attested cardinality (accumulator index)
│
├─ 41 legacy MIC tranche      ← C-299–C-307 (indexed, outside collision audit scope) ✅ classified step 2
│
└─ 319 modern Reserve Block   ← C-308+ attested bodies in C-377 audit / collision witness scope
    ├─ 194 candidate positions ← deduped unique block_number (NOT constitutional canon)
    └─ 125 collision pairs / 123 contested

The 41-ID gap is classified (Track R step 2, 2026-08-08). Do not publish canonical_reserve_blocks = 194 or any other fixed count until step 8. The API field must remain unresolved.

Future Terminal UI (deferred): surface 360 INDEXED · 319 EXAMINED · 194 CANDIDATE POSITIONS · CANON UNRESOLVED so operators cannot confuse index cardinality with constitutional canon.

What the C-368/C-397 cold snapshot candidate proves¶

The 194-position candidate chain (e.g. PR #419 replay) is internally cryptographically valid when re-verified:

• 194 contiguous candidate positions
• 9,700 MIC at 50 MIC per position
• both .dat SHA-256 digests match MANIFEST.json
• every prev_hash link and per-block hash verifies
• chain tip matches the manifest

That proves artifact integrity of a candidate export. It does not adjudicate which competing historical seal belongs in each of the 123 contested positions, and it does not resolve the 41-ID index/examination gap.

Track R recovery gate (ordered)¶

C-397 does not ship a regenerated .dat candidate in this tree—the canon rollback left checked-in canon/reserve-blocks/ unchanged from main. What is preserved here is verification evidence only: the 125-pair witness JSON, the C-377 audit provenance, and witness-table rows for #419 candidate replay hashes (not retrievable as Substrate artifacts after rollback).

Merge-ready sequence — do not skip or reorder without recording why:

1. Fresh collision audit — production KV; confirm pair count still matches witness (or regenerate evidence).
2. Classify the 41 index/examination-gap IDs — ✅ done (2026-08-08): legacy MIC tranche C-299–C-307; see gap witness JSON.
3. Confirm live unique-position candidate count — verify whether deduped candidate positions remain 194 after step 1–2 (may change).
4. Adjudicate the 123 known contested positions — Track R receipts per position; no silent dedupe winners.
5. Human + ZEUS + EVE approval — recorded on PR / consensus gate.
6. Apply guarded Track R repair — C-373 collision-repair transaction only; no KV rewrite outside guard.
7. Post-repair audit — zero unresolved hash-divergent divergence; SEAL_INTEGRITY_GATE may lift only after this.
8. Determine canonical RB count and regenerate .dat — only after steps 1–7; then canonical_reserve_blocks resolves from Track R evidence, not from index cardinality or candidate dedupe alone.

Invariant constraints (all steps):

1. Preserve all original seal bodies.
2. Do not renumber or delete competing seals.
3. Do not disengage SEAL_INTEGRITY_GATE until live audit and canonical-count evidence resolve.
4. canonical_reserve_blocks stays unresolved until step 8 — never preset to 194.

Track R step 1 — fresh production audit (2026-08-08) ✅¶

Manual workflow Audit Reserve Block Lineage #6 on mobius-civic-ai-terminal@main:

Step 1 verdict: collision witness still valid — no regeneration of C397_RESERVE_BLOCK_COLLISION_WITNESS.json required for pair counts.

Lineage audit adds structure (artifact lineage-audit.json, same run):

The four lineage components (by tip):

1. C-332-001 → C-358-131 — seq 1–131 (131 seals)
2. C-359-001 → C-371-033 — seq 1–33 (33 seals)
3. C-372-001 → C-372-002 — seq 1–2 (2 seals)
4. C-308-042 → C-332-194 — seq 42–194 (153 seals; orphan link at genesis of this fork)

Still open after step 1:

• Step 3: candidate count still 194 among examined set — unchanged, still not constitutional canon.
• Steps 4–8: unchanged.

Track R step 2 — index/examination gap classified (2026-08-08) ✅¶

Operator KV export (vault:seals:index, :attested, :all, vault:seal:latest) proves the 41-ID gap is not a mystery subset — it is exactly the legacy MIC tranche era:

Arithmetic: 360 indexed − 41 legacy = 319 examined — the gap closes without orphan or missing-body IDs.

Evidence artifact: C397_INDEX_EXAMINATION_GAP_WITNESS.json

KV observations (same export):

• vault:seals:index, :attested, and :all are byte-identical (360 IDs, no quarantine delta).
• vault:seal:latest points to seal-C-372-002 (C-372 fork tip) — an accumulator/migration pointer from C-305 v1 migrate, not canonical_reserve_blocks tip while SEAL_INTEGRITY_GATE is active.
• Zero legacy-tranche IDs appear in C397_RESERVE_BLOCK_COLLISION_WITNESS.json collision pairs — they are not collision candidates.

Step 2 verdict: classify as legacy_mic_tranche_outside_audit_scope. Do not adjudicate these 41 as contested Reserve Block positions; Track R steps 4–8 apply only to the 319 modern-era seals (125 pairs / 123 contested positions among them).

Authority declared using docs/templates/EPICON_FOUNDER_STANDING.md v0.1

C-397 invokes founder–custodian standing only to preserve reconciliation evidence and hold the Reserve Block canon gate fail-closed while lineage is contested.

Scope Constraints¶

This authority is narrowly scoped to:

• preserving the C-377 pair-count witness,
• preserving the verification facts for the #419 candidate replay (hashes recorded in Witness Table—not a checked-in artifact),
• requiring Track R evidence before canonical promotion,
• requesting independent sentinel and human review.

It does not authorize production KV mutation, seal deletion/rewrite/renumbering, gate disablement, MIC issuance changes, or unilateral selection of canonical winners.

Temporality & Revocation¶

This authority is transitional, contestable, and non-transferable. It may be superseded or revoked by a successor EPICON, ratified governance process, or contrary reconciliation evidence.

Legitimacy Rationale¶

The integrity hold affects consequential ledger state. Making the custodian action explicit is preferable to allowing an implicit operator choice to decide which historical lineage becomes canon.

Acknowledgement of Risk¶

The proposer acknowledges that founder authority is asymmetric and that the deterministic export winner is not automatically the legitimate historical winner. Future governance may revise this decision while preserving the evidence trail.

Sunset Condition¶

This C-397 authority expires for this incident when either:

1. Track R reconciliation is approved and canonical-count evidence is published, or
2. a successor governance/EPICON decision supersedes C-397.

The separate intent timebox in PR #429 remains an outer bound.

Witness Table¶

Restraint row¶

• No production KV mutation in this PR.
• No canonical .dat or manifest change before Track R adjudication.
• No SEAL_INTEGRITY_GATE disable.
• No historical seal deletion, rewrite, or renumber.
• No MIC issuance or conversion change.
• No preset canonical_reserve_blocks = 194 (or any fixed count) before Track R step 8.
• No UI-derived canonical count.
• No claim that cryptographic chain validity substitutes for lineage adjudication.

One truth, three skins. Canon → Ledger → UI.