R19: Run two disposable OpenClaw-scale canaries

[100yenadmin]

Use a disposable copy, never the live index or its only copy. Record repository SHA, graph counts, embeddings, model, dimensions, sidecars, disk space, and dirty state. Run clean analysis, controlled incremental changes, forced escalation, interruption at durable phases, and resume; repeat the successful sequence twice.

Acceptance: two complete runs with no unexplained count change, secret leakage, silent loss, unsafe recovery, native crash, or false-success exit.

Validation: retained secret-free transcripts and before/after state manifests.

Parent epic: #39

Tracking ID: R19

[100yenadmin]

Current exact next action (2026-07-13): OpenClaw canary 1 is active on disposable clone /Volumes/LEXAR/repos/gitnexus-r19-openclaw-canary-1, source c0465b863c798e35cc7fac882e1a2c5583b4b21b. Clean analysis, controlled add, indexed embedding deletion, controlled edit/rename/delete/add, FTS query, context, and impact have passed. The 51-file importer-closure phase proved that 10-minute and 45-minute harness ceilings were invalid at this scale while remaining pre-mutation with no dirty marker. The resumable harness now uses a configurable six-hour ceiling; timing failures are preserved separately. No live index is involved.

[100yenadmin]

R19 OpenClaw canary 1 graph mismatch

Date: 2026-07-14 (Asia/Bangkok)

Identity

• GitNexus branch/head used by the completed run: reconcile/r19-openclaw-canaries at ea1f59a233a712ceaa995453537bc5b62ce98415.
• Disposable OpenClaw source: c0465b863c798e35cc7fac882e1a2c5583b4b21b plus canary-only fixture commits; final disposable head b4d02e675f521c9eeeaff90e39470d6abaa241e7.
• Evidence: r19/run-1/.
• Disposable repository: /Volumes/LEXAR/repos/gitnexus-r19-openclaw-canary-1.
• Embedding provider: local deterministic test server, model gitnexus-r19-deterministic-v1, dimension 8.

Proven behavior

• Clean analysis, controlled add, edit, rename, delete, importer closure, FTS query, context, impact, forced escalation, SIGTERM interruption, dirty-marker detection, recovery, and a forced rebuild all completed without a native crash or false-success exit.
• Recovered and forced states both retained 251532 queryable embeddings at dimension 8.
• FTS and vector-index capabilities were available after both successful analyses.
• The live OpenClaw index and its only copy were never used or mutated.

Release blocker

The recovered graph and the immediately following forced rebuild used the same disposable source commit but did not produce identical graph counts.

Metric	Recovered	Forced	Delta
Files	20130	20130	0
Nodes	268389	268400	+11
Edges	1185856	1185667	-189
Communities	15423	15436	+13
Processes	300	300	0
Embeddings	251532	251532	0

The completed run stopped non-zero with recovered/forced nodes count mismatch. This is a release-signoff blocker, not a blocker for unrelated sprint work.

Diagnostic follow-up

The canary harness now records per-node-table counts, stable SHA-256 digests of sorted node IDs, relationship-type counts, and relationship label-pair counts. A second forced rebuild is running from the existing disposable forced state:

• Matching consecutive forced fingerprints will isolate the discrepancy to interrupted recovery or its inputs.
• Different consecutive forced fingerprints will demonstrate broader full-build nondeterminism.

No upstream issue will be opened until the result is reduced to an actionable, sanitized reproduction.

[evaos-code-review-bot]

evaOS issue enrichment

Issue: #58 - R19: Run two disposable OpenClaw-scale canaries
State: open; milestone: GitNexus v1.6.10 Reconciliation - Sprint 3 - Retrieval Resilience and Scale Proof

Related issues/PRs: #39.
Existing labels: reconciliation, P0, evidence-gate, release-blocker.
Suggested labels: security, support.
Suggested owners: none.

Related context

• EPIC: Reconcile evaOS GitNexus with upstream v1.6.10 #39 - mentioned in issue metadata; inspect for dependency, duplicate, or prior-decision risk.

Agent-start packet

Problem shape: R19: Run two disposable OpenClaw-scale canaries - Use a disposable copy, never the live index or its only copy. Record repository SHA, graph counts, embeddings, model, dimensions, sidecars, disk space, and dirty state. Run clean analysis, controlled incremental
Product fit: Implementation issue; keep product fit lightweight unless VISION.md/repo policy raises a constraint.
Implementation wedge: Start with a focused fixture or dry-run path that proves the issue shape before widening behavior.

Build / borrow / buy scan:

• Build: identify the smallest implementation wedge that proves the issue without broad rollout.
• Borrow: inspect same-repo prior art, repo memory, GitNexus context, and allowlisted cross-repo examples first.
• Buy/use: deferred; no product, architecture, UX, library-choice, integration, roadmap, or market-positioning trigger detected.

Candidate sources:

• VISION.md / repo policy / repo-memory.md / GitNexus packet.
• Same-repo issues and PRs referenced by issue metadata.

Acceptance criteria:

• Preserve and execute the issue's stated acceptance criteria.
• Include source citations for any external OSS/library/current-market claims.
• Keep issue-enrichment comments sticky and idempotent.
• Keep labels, owners, reviewers, and roadmap fields suggestion-only.

Proof plan:

• Focused unit/fixture test for the changed planner behavior.
• Dry-run issue-enrichment evidence packet before live comment posting.
• Redaction check for issue body, citations, and generated markdown.

Known traps:

• Do not treat related links as proof by themselves; each source needs a reason why it matters.
• Do not scan old issue backlogs or exceed repo-level enrichment throttles.

Non-goals:

• Do not auto-apply labels, owners, reviewers, roadmap fields, or milestones.
• Do not bulk-enrich old backlog issues from this planner output.
• Do not claim external market or OSS research was performed unless cited sources are present.

Context-source taxonomy:

• [enabled] vision_repo_policy_memory_gitnexus: Use repo vision, policy, memory, and GitNexus first before external research.
• [enabled] same_repo_issues_prs: Issue metadata contains same-repo references.
• [deferred] allowlisted_cross_repo_github: Deferred until product, UX, market-positioning, or library-choice signal appears.
• [deferred] external_oss_examples: Deferred for low-research-value issue class.
• [deferred] library_api_docs: Deferred unless a library/API decision is present.
• [deferred] current_market_examples: Deferred for implementation-only issues.
• [enabled] internal_mcp_docs: Use internal docs/MCP context when available; missing context is degraded, not blocking.

Validation suggestions

• Confirm owner, acceptance criteria, and validation evidence before implementation.

Triage gaps

• Owner or reviewer signal not detected in issue body.

No labels, owners, reviewers, or roadmap fields were changed by this bot.

[100yenadmin]

Disposable OpenClaw canary update:\n\nThe first full canary completed clean analysis, controlled incremental changes, forced escalation, interruption/resume, FTS/vector checks, and preserved 251,532 embeddings. The final consecutive-force equivalence gate correctly failed.\n\nAcross two forced rebuilds of the same source commit:\n- source node IDs and counts were stable for every non-Community node table;\n- embeddings remained exactly 251,532;\n- ACCESSES changed by +60;\n- CALLS changed by +112;\n- MEMBER_OF changed by +16;\n- Community count changed from 13,733 to 13,712.\n\nThis is a release-signoff blocker, but it is not holding unrelated sprint work. A third isolated rebuild is running with bounded SHA-256 relationship identity fingerprints (endpoints, labels, confidence, reason, and step) to distinguish insertion duplication from order-sensitive relationship resolution. No live index is involved.

[100yenadmin]

R19 implementation artifact is now open as PR #92, stacked on the current reconciliation head.

Focused proof before opening:

• harness and subprocess child pass node --check;
• both files pass Prettier;
• large-incremental-subprocess-e2e.test.ts passed 1/1 in 248.25s;
• the active disposable canary remains isolated from the live OpenClaw index.

The PR does not claim R19 acceptance. Issue closure remains gated on resolving the graph-fingerprint discrepancy and completing two successful disposable OpenClaw-scale sequences. Exact next action: finish the current relationship-identity repeat, classify the mismatch, apply any proven fix on the stacked branch, then run two clean final-head canaries.