GLM runtime investigation evidence¶
This bundle supports the dated finding. The summary is the authoritative source for displayed results; coverage lists the unrun gates. Three matched DCP1 175K/32K overlaps passed after the DCP2 parent crashed. This is bounded mitigation evidence, not kernel attribution, a reliability rate, performance ranking or replacement qualification. The exact baseline is restored.
Artifact manifest binds every retained byte; manifest source reproduces it. Native stability artifacts retain failures, usage, response IDs and bounded outputs as originally collected. The initial parent and 512-cap runs predate the result-retention fixes; see friction and correlation.
Sanitization receipts bind private source hashes to public copies. Generic paths, ports and GPU identities are reconstruction placeholders. Container IDs are consistently redacted with one-way tokens; the evidence inspector marks their provenance as sanitized. Recipe/registry digests inside native evidence identify the original private configuration; recalculate them when loading a local copy. The endpoint placeholders must not be used as proof that a recipe is running.
No graph or social-copy package is required: these are diagnostics without comparable performance populations. All files are text; raw evidence stays below 1 MiB per file and 5 MiB for this bundle.
Source registry¶
Workload manifest¶
Run plan¶
Configuration and identity¶
- baseline-recipe.toml
- candidate-dcp1-recipe.toml
- baseline-recipe-status.json
- baseline-gpus.csv
- dcp1-start-identity.json
- dcp1-start-status.json
- dcp1-admission.json
- dcp1-startup.log
- development-source.json
- sanitization-receipts.json
- dcp1-small-scenario-comparison.json
- dcp1-175k-scenario-comparison.json
Raw run evidence¶
- baseline-small.json
- baseline-175k-serial.json
- baseline-175k-overlap.json
- baseline-preflight.json
- baseline-preflight-thinking-enabled.json
- baseline-json-thinking-enabled.json
- json-policy-disabled-2.json
- json-policy-disabled-3.json
- json-policy-enabled-2.json
- json-policy-enabled-3.json
- dcp1-preflight.json
- dcp1-small.json
- dcp1-small-matched.json
- dcp1-175k-overlap.json
- dcp1-175k-repeat2.json
- dcp1-final-metrics.json
- dcp1-final-owner.log
- dcp1-final-status.json
- dcp1-final-kernel.log
Failures and friction¶
- friction-log.md
- crash-175k-correlation.json
- crash-175k-overlap-model-combined.log
- crash-175k-overlap-kernel.log
- crash-175k-overlap-status.json
- dcp1-small-kernel.log
- dcp1-small-owner.log
- dcp1-201k-time-admission.json
- skill-review-report.txt
- skill-review-operator-workflow-v1.json
- skill-delta-review.json
- skill-scenario-comparability-review.json
- runner-review-findings.json
- runner-review-operator-workflow.json
- dcp1-followup-findings.json
- dcp1-followup-operator-workflow.json
- dcp1-result-retention-followup.json
- dcp1-result-retention-operator-workflow.json
- dcp1-continuation-review.json
- dcp1-continuation-operator-workflow.json
Restoration¶
- restoration.json
- restoration-status.json
- restoration-startup.log
- restoration-direct-preflight.json
- restoration-routed-preflight.json
- restoration-router-sha256.json
- restoration-kernel.log
- restoration-gpus.csv
- recovery1-preflight.json
- recovery1-routed-preflight.json
Decision summary¶
Independent final source and publication review accepted the reviewed bytes for exact PR-head CI. Its manifest hash deliberately records the bundle before insertion of this review packet.