[yifjiang]

Summary

Owned prompt copies, deadlines, blocking-pool saturation and departure handling; query-only paths never book decode capacity.

Review slice D11 of #16748; part of #16747 and #16746. This draft targets codex/kv-reuse-d10-source-eligibility for a complete incremental review diff of 485 additions plus deletions, including tests, documentation and extraction glue.

Extraction details:

• Host and GPU candidate helper methods move here with their first production caller. Asynchronous query/deadline/semaphore ownership and all CPU lookup regressions are preserved.

Reuse remains experimental and default off. Retained inventory is advisory, complete compatible rank parts must precede readiness, and native operations retain ownership until quiescence. Supported GPU activation also requires the companion KVCR/native slices and the declared deployment profile.

The lower-tier lookup helpers compile with their optional standalone-selection consumers. D12 additionally exposes the GPU helpers to their default unit tests; no test function is gated or skipped by this correction.

Validation

All configured pre-commit hooks passed for this exact slice on allocated SLURM. Stack-level CPU regression coverage is summarized on D31; each draft still requires its own applicable CI in the Checks tab. Historical combined-head and hardware results retain their original source scope.

[devin-ai-integration]

✅ Dynamo PR CI passed — run 41941643671 (attempt 2) on 7e7acd1c3c

Gate checks: ✅ backend-status-check · ✅ deploy-status-check · ✅ dynamo-status-check

_{Posted automatically by Devin for run 41941643671. Updated on every full-CI run of this PR.}

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[yifjiang]

/ok to test fded18a

[yifjiang]

/ok to test 7e7acd1