fix(pair): pair_await waits the whole cap so a lead cannot poll

[rynfar]

What

pair_await no longer lets the lead pick a short wait. Any request except an explicit 0 waits the lead's whole cap (45 s, or 150 s where Pylon sets the provider's tool timeout). 0 still reads the state and returns at once.

Why

In a live run a Codex lead called pair_await with maxSeconds: 10 and 20 in a loop, twelve calls in four minutes against an executor that never finished. Each return is a model turn, which is the polling cost pair mode exists to remove. The call already returns the moment the executor stops or needs the user, so a long wait loses nothing.

The old default was also wrong: leaving maxSeconds out meant "do not wait", so the most natural call returned instantly and invited a loop.

How

• pairAwaitBudgetSeconds(requested, cap) in pair/logic.ts: 0 gives 0, anything else gives the cap. The handler uses it in place of Math.min(requested ?? 0, cap).
• The parameter description now says to leave it out, that values above 0 are raised, and that 0 is a state read.
• Two handler tests that encoded "omitted means no wait" now ask for maxSeconds: 0.

Verification

• vp test run apps/server/src/mcp/toolkits/pair: 55 passed, including the new rule.
• Typecheck t3: 0 errors. Format check clean.
• Not run live after the change.

Known limits

• A lead can still loop on maxSeconds: 0. Nothing counts repeated reads yet.
• delegated_thread_status still honors short waits; the fan-out tools are being superseded by pair mode, so they are left alone here.

Part of #622.

Model: Claude Fable 5.1, 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 5:40pm 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	13.9 KiB	−13 B (−0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.2 KiB	7.2 KiB	0 B (0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.7 KiB	6.7 KiB	−13 B (−0.2%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	58.0 KiB	58.0 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	9	9	0 (0.0%)	21	✅
Claude	Total thread wire	14.0 KiB	14.0 KiB	0 B (0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.2 KiB	7.2 KiB	−2 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.7 KiB	6.7 KiB	+2 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	58.9 KiB	58.9 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 6468f56 · PR result: c2f2084 · 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: 115.7 KiB
• Claude decoded thread snapshot: 116.3 KiB

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