fix(relay): sweep expired state hourly instead of every five minutes

[rynfar]

What Changed

infra/relay/src/worker.ts: the housekeeping cron goes from */5 * * * * to 0 * * * *, with a comment explaining why. It still does exactly the same two things — delete expired DPoP replay rows and delete terminal agent-activity rows older than 30 minutes.

Why

Neon's Free plan suspends the compute after five idle minutes and allows 100 CU-hours a month. A sweep every five minutes means the compute is woken 288 times a day and effectively never sleeps: 0.25 CU × 24 h × 30 d ≈ 180 CU-hours/month at the minimum size. The project ran dry on 2026-08-26 and the relay went down with database_unavailable.

Hourly costs ~24 short wakeups a day ≈ 15 CU-hours/month for housekeeping, leaving the budget for actual use.

Nothing depends on the sweep's timing:

• DPoP replay is an INSERT … ON CONFLICT DO NOTHING on the proof id (auth/DpopProofs.ts:56-79); a reused id conflicts whether or not the expired row has been swept. Proofs expire after 5 minutes, so the table holds at most an hour of a single user's traffic.
• Terminal Live Activity rows stop rendering after TERMINAL_AGENT_ACTIVITY_DISPLAY_TTL_MS (15 min, AgentActivityPublisher.ts:232) and are excluded from live-work checks by isTerminalPhase/isExpiredAgentActivityState — all based on updatedAt, not on whether the row still exists. The sweep only reclaims space.

Not in this PR (console settings, not code): pin the Neon compute to a fixed 0.25 CU so bursts can't autoscale to 2 CU.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes (no UI change)
• I included a video for animation/interaction changes (no motion change)

Model: Claude Fable 5 · Harness: Claude Code

---

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

[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	13.3 KiB	13.2 KiB	−100 B (−0.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+8 B (+0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.3 KiB	−108 B (−1.7%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.5 KiB	54.7 KiB	−822 B (−1.4%)	66.4 KiB	✅
Codex	Live turn messages	9	9	0 (0.0%)	21	✅
Claude	Total thread wire	13.3 KiB	13.3 KiB	+3 B (+0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+3 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	0 B (0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	56.3 KiB	56.3 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 486065e · PR result: e0ff0a3 · 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: 109.4 KiB
• Claude decoded thread snapshot: 110.0 KiB

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