fix(server): back off crashing tunnel connectors

[rynfar]

A relay connector that exits immediately can restart in a tight loop and exhaust the Pylon server’s memory, disrupting every connected agent. Back off repeated rapid exits while keeping the first restart immediate; reset the delay after stable uptime or an explicit configuration change. Configuration changes can preempt a delayed restart.

Adopts approved B1, upstream #8788, b90898077e60290a1eb7bea4224f325fd3bb1595. Pylon’s process ownership, runtime home, branding and configuration routing are preserved. This affects the managed tunnel connector for every client/provider; direct local connections are unchanged.

Validation: 11 focused ManagedEndpointRuntime tests pass, including crash-loop delay, stable-uptime reset, configuration replacement and existing serialization/cleanup cases. Server typecheck, targeted lint and diff checks pass. Tests use fake child handles and Effect TestClock; no real tunnel or provider was launched.

Model: GPT-6 Astra. Harness: Codex.

---

^{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	13.6 KiB	13.6 KiB	−6 B (−0.0%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	−5 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	−1 B (−0.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.5 KiB	56.5 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	8	8	0 (0.0%)	21	✅
Claude	Total thread wire	13.6 KiB	13.6 KiB	+1 B (+0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	+4 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	−3 B (−0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.3 KiB	57.3 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	8	8	0 (0.0%)	21	✅

Baseline: c3fbc15 · PR result: ba62757 · 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: 111.2 KiB
• Claude decoded thread snapshot: 111.9 KiB

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