Runtime scout evidence

The finding records two separate private-target experiments. The original configuration was retained after each one. This public bundle is a sanitized numerical record, not a complete reproduction package.

• Artifact inventory and hashes
• Sources
• Configuration, redactions and reproduction limits
• Decision summary
• Restoration
• Failures and limitations
• Cache-layout summary, plan excerpt, chart, chart data
• DFlash summary, plan excerpt, chart, chart data

Each experiment retains twelve timed native JSON artifacts per arm, named baseline-before-00.json through -11.json, candidate-00.json through -11.json, and baseline-after-00.json through -11.json. Each has its request journal, counter reconciliation and sampled telemetry. Allocation and preflight files are separate. The manifest inventories every file. Numeric fields are preserved; the same sanitized placeholder does not imply that distinct private arms had identical configuration fingerprints.

The source bundles were hash-verified before copying. Protected originals were left unchanged. Public source hashes refer to these sanitized bytes. The publication validation recomputes pooled timing statistics from retained native request rows and verifies allocation and restoration receipts.