[novelKR]

Before: the ledger records P1-RECOVERY as in-progress, qualification not-run, with P1R-C06 active, and the plan says the new mode is not a supported mode until P1R-C06 qualifies it.

After: the ledger records P1-RECOVERY as implemented, qualification macos-qualified, with its qualified_combination and every unit merged-verified. The dated notes, P1-RECOVERY's rows, V-P1 and the R3 gate name the qualification and cite CodeSpace's report (novelKR/CodeSpace#116).

This is the last step of P1-RECOVERY's re-baselined plan, the same kind of change as #35 for CS-RG. Tracker: novelKR/CodeSpace#101.

Why

On 2026-10-14 the owner directed P1-RECOVERY in CodeSpace as one pull request per unit, P1R-C01 to P1R-C06, as re-baselined in its plan (#36). P1R-C01 to P1R-C05 are merged and verified (#37 to #41). P1R-C06's harness, novelKR/CodeSpace#112, and its three harness fixes, novelKR/CodeSpace#113, novelKR/CodeSpace#114 and novelKR/CodeSpace#115, are merged. The candidate 1c0eadd then passed every suite of CodeSpace's frozen protocol on the qualification host, and the report merged as CodeSpace main d0f2e49, documentation only, with CI green. The ledger still records P1-RECOVERY as in-progress, qualification not-run, with P1R-C06 active.

Result

Item	Result
Combination	CodeSpace 1c0eadd, code tree cb18534d…8d38; DevGuard's client 6e7e065; release 0.1.0-6e7e065-94d85365, with DG-1's record fd4fed31…17a4, under the policy CS-RG was qualified with (configuration fingerprint 8cbb019b…fc6f); Codex 6b9826e with CodeSpace's recorded patch, patched tree 1999937; Rust 1.95.0
Qualified	uds with --runner-lifecycle independent and a Runner the operator starts, with pipe and PTY executions, participations off and required, and both in one gateway
Not qualified	Linux off is CI functional verification only; Linux required stays unsupported until DG-LINUX; restoring processes after the Runner itself restarts is unsupported and outside P1-RECOVERY
Functional suites	55 of 55 scenarios that run on macOS passed, 34 with off and 21 with required, with 650 checks, three times: in the five check suites, the rehearsal and the run. The protocol's two gaps were recorded as not run
Restart measurement	qualified: three repetitions of an idle and a load interval under DG-1's cold load, all six intervals passed, 308 Gateway restarts in all; worst status p99 3.283 ms (target 500 ms) and worst termination acknowledgement p99 0.739 ms (target 1,000 ms)
Recovery	after every restart the next gateway found each long-running process running; no command ran twice; all 4,710 requests from another process were refused STALE_CONTROLLER and changed nothing; no timeout was extended, and the latest ended 119 ms after its deadline (limit 2 s); no output line was lost or repeated outside a gap the Runner declared
Regressions	CS-RG's candidate, matrix (45 of 45), faults (27 of 27) and authority (10 of 10) passed again on the candidate; CI passed every job on 1c0eadd, in the push run and the scheduled full run
Evidence	sealed on the host and listed with each MANIFEST.json SHA-256 in the report; verify accepted all 35 P1R-C06 directories, and a scan of 13,454 files found no credential

Changes

File	Change
milestones.json	P1-RECOVERY implemented, qualification macos-qualified, with its qualified_combination: CS-RG's keys, plus runner_lifecycle independent; units as_of 2026-10-19 and P1R-C06 merged-verified
AGENTS.md, docs/milestones.md	The current-state note, and P1-RECOVERY's row names the qualification, its pull requests and the report
docs/planning/README.md, Korean pair	The status sentence, the current-state note and P1-RECOVERY's row
docs/planning/milestones/P1-RECOVERY.md, Korean pair	The dated current-state note records the qualification, its combination and what stays unsupported
docs/planning/verification.md, Korean pair	V-P1 names the qualification instead of "Future"
docs/planning/consumer-readiness.md, Korean pair	The independent Runner mode joins R3's qualified modes for its combination on that host

Korean counterparts reviewed; the planning-readme, planning-consumer-readiness, planning-milestones-p1-recovery and planning-verification hashes recorded in docs/translations.json with scripts/check_docs.py record.

What does not change

• The independent mode stays opt-in. Making it the default is a separate decision, as ADR-002 says.
• CS-RG's qualification and its combination ca306ee; DG-1's qualification and release.
• DG-LINUX, DG-CACHE and DG-ADAPTERS stay not-started.
• The suspension of the CS-RG plan's CSRG-C units, and every historical text.
• No product code, contract, pin, dependency, release, CI policy or test change.

Verification

At 4044575, based on main at ec66b12:

• python3 scripts/check_docs.py: passed, 18 reviewed pairs, 48 work units, 25 logical PR groups.
• python3 scripts/validate.py --stages whitespace,source-contract,documentation --diff-base origin/main: every stage passed.
• python3 scripts/ci_plan.py --base origin/main: profile affected, 12 normative paths and no component, so the repository stages whitespace, source-contract and documentation run and Rust does not.
• python3 -m unittest discover -s scripts -p 'test_ci_*.py': 95 tests passed.
• milestones.json parses, and its qualified_combination matches CodeSpace 1c0eadd, its Codex gitlink 6b9826e3 and DevGuard pin 6e7e065, and the patched tree that CodeSpace's report records.
• CI on this head: run 39198848429 passed the plan, the repository checks and the required checks; the plan skipped contracts, as nothing outside the documentation and the ledger changed.

Rollback

Revert the merge (git revert -m 1 <merge commit>). The ledger then records P1-RECOVERY as in progress again, while CodeSpace's report stays the record of the runs.