feat(pair): delete executors left behind by abandoned drafts

[rynfar]

What

When the server starts, it deletes pair executors that nothing will ever brief.

Why

Turning Pair on in a draft creates the executor before the lead thread exists, because the first send creates the lead, its worktree and its first turn in one step. If the draft is abandoned, or its id changes before that first send, the executor stays forever as an empty "Executor · New thread" in the delegated group.

How

• orphanedExecutorIds (pure) is deliberately conservative. A missing lead proves nothing on its own, since that is what every freshly paired draft looks like. An executor counts only if it is a pair executor, its lead is not among the active or archived threads, it never ran (no turn, no session), and it is more than a day old.
• PairLifecycleReactor.start runs the sweep once before subscribing to events and dispatches thread.delete with a deterministic command id, so a restart replays harmlessly. A failed sweep logs a warning and never stops the reactor from starting.

Verification

• vp test run on apps/server/src/orchestration and apps/server/src/mcp/toolkits: 957 passed. New, written first and seen failing: the pure rule with six "leave it alone" cases (existing lead, fresh draft, an executor that ran, one with a session, a fan-out child, an ordinary thread), and a reactor test where only the abandoned draft's executor is deleted while an executor whose lead is archived is kept.
• Typecheck t3: 0 errors. Lint, format and vp run knip:check clean.
• Not run against a real database.

How it was built

Second change made through pair mode on a real install: Claude lead, Antigravity executor, one shared worktree, tests protected (protectedPaths: checked 2, changed []). One lead correction after review: the executor had made production code fall back to the snapshot time when the clock read zero, to get past a test running under a test clock. The test now sets the clock and the reactor reads it plainly.

Part of #622.

Model: Claude Fable 5.1 (lead) with an Antigravity executor, paired in Pylon via the Claude Code harness.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
pylon-marketing	Ready	Preview	Sep 18, 2026 10:12pm UTC

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	14.0 KiB	14.0 KiB	+5 B (+0.0%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.2 KiB	7.2 KiB	−6 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.7 KiB	6.7 KiB	+11 B (+0.2%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.0 KiB	58.0 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	14.0 KiB	14.0 KiB	−4 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.2 KiB	7.2 KiB	−18 B (−0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.7 KiB	6.8 KiB	+14 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.9 KiB	58.9 KiB	+44 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	9	10	+1 (+11.1%)	21	✅

Baseline: 16c373e · PR result: 098f6e9 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 115.6 KiB
• Claude decoded thread snapshot: 116.3 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.