Session relaunch: base completion on the final live list; a registered-then-exited session is absent (#1116 follow-up a) - #1140

Scope

• What this PR changes, in one or two sentences: Session relaunch (Session relaunch after an unplanned WSL restart: a roster timer and a boot oneshot on the systemd user manager #1116) now decides completion from the run's final live list. A session seen running earlier in a run and missing at its end is absent, so the run is partial and a rerun brings it back. An unreadable final list never completes a run.
• Base commit: 0f4ee075b
• Lane: lane:foundation
• Owned paths touched:
  • tools/session-relaunch/session_relaunch.py and tools/session-relaunch/README.md;
  • tests/test_session_relaunch.py;
  • docs/decisions/2026-10-14-session-relaunch-after-wsl-restart.md (the tests line);
  • evidence/artifacts/session-relaunch-20261014/README.md (the test count);
  • manifests/evidence.json.

This is #1116 follow-up (a), the first in the command center's order (2026-10-15T02:55Z): (a) completion, then (b) the roster lock and (c) the carry after a partial run, then (d) the P3s.

Finding: the local review's P2 at 7aad5a0, line 533; at c8d7fa9 it is the final pass, lines 547-553. Source: the local review, review_output.json under /var/tmp/ns2604-local-review-20261013/.

• An early-started lane registers, then exits while later tabs open.
• Its earlier running outcome stood.
• The run said complete, and plan() refused every later rerun as already relaunched.

Fix:

• In the final pass, a session seen running earlier and missing from the final list becomes absent.
• Its resolved_through moves past its calls, because its tab start was seen to register. A rerun therefore opens it again without a --reopen.
• unresolved() is unchanged for a session never seen listed, so the double-launch closure of the earlier rounds holds.
• complete now requires a final list that was read. An unreadable one leaves the run partial (exit 1, the alert), and a rerun reads it again.

SOTA sources

• Code under change: native-agent-stack/native-agent-stack at c8d7fa9e42c62bffc24446e7829ecc5933ec945d, tools/session-relaunch/session_relaunch.py:547-553 (the final pass) and :320-327 (unresolved()).
• Live list: the final pass reads the live list from claude agents --json, the client's own session list. Its shape was measured at Claude Code 2.1.299, in Session relaunch after an unplanned WSL restart: a roster timer and a boot oneshot on the systemd user manager #1116's evidence/artifacts/session-relaunch-20261014/receipts/probes.json (:4; client also at sources.json:3).
• Upstream: own logic; no upstream mechanism changes.

Evidence-class table

Local commands run

$ python3 -m unittest tests.test_session_relaunch -k registers_and_then_exits -k unreadable_final_list   # at c8d7fa9e
FAILED (failures=2)   # 'complete' != 'partial', both
$ python3 -m unittest tests.test_session_relaunch
Ran 53 tests; OK
$ python3 scripts/validate.py
rc 0; {"components": 71, "hashed_files": 12040, "profiles": 4, "receipts": 240, "status": "passed"}

Decision record

docs/decisions/2026-10-14-session-relaunch-after-wsl-restart.md: its tests line gains this follow-up and the count of 53.

Follow-ups

• (b) and (c) follow in one PR, since they share the snapshot's carry. (b) adds a short roster lock around the seed roster write and the timer snapshot's carry. (c) carries a partial run's unfinished rows into this boot's roster.
• (d) covers the P3s of the GPT fix3 read and the command center's micro, and the README's dry-run wording: the dry run writes relaunch.lock.

Checklist

• No GitHub Actions changed.
• No workflows changed.
• No secrets are printed, logged or committed.
• No new paid hosting, subscription or billing surface.
• Peer-owned untracked files and worktrees were preserved.

🤖 Generated with Claude Code