MC-4735

[follow-up] 20 min past expected check-in (2026-06-04T21:41:02.059584+00:00 UTC).

[control-room-promote] Promoted MC-4735 from inbox to todo (priority medium) so the Control Room pickup loop can dispatch it. Deterministic triage — no worker fired yet.

[control-room-dispatch] Control Room dispatched MC-4735 to a Claude Code worker. Worktree: /home/lucienne/workspace/state/control-room-worktrees/mc-4735-update-hermes-agent-on-luci-384-commits-112763 Branch: cr/mc-4735-update-hermes-agent-on-luci-384-commits-112763 tmux: cr-MC-4735 Expected check-in: 2026-06-04T21:41:02.059584+00:00

[follow-up] 562 min past expected check-in (2026-06-04T21:41:02.059584+00:00 UTC).

RUNTIME TERMINAL STATE (MC-3482 contract) status: warning summary: Live ticket runtime timed out without a harvestable verdict; parked for automatic runtime recovery. root_cause: harvest timeout — pane was silent/wedged and the transcript carried no verdict to sweep safe_retry: Runtime was killed and the ticket was returned to todo for a fresh Luci-owned retry; no human input is needed. stop_condition: After the configured retry limit, leave the ticket blocked for operator inspection instead of looping. human_input_required: no next_actions: - Pickup will start a fresh runtime from MC history. - Inspect the pane log artifact only if the retry fails again. artifacts: - ticket:MC-4735

Operator note: Elmar asked what input is needed. Current state: no product/user decision needed; stale worker from 2026-06-04 is silent/wedged. Need only a short maintenance window because Hermes update/restart may briefly interrupt Telegram. I am clearing the wedged runtime so pickup can retry cleanly.

[control-room-dispatch] Control Room dispatched MC-4735 to a Claude Code worker. Worktree: /home/lucienne/workspace/state/control-room-worktrees/mc-4735-update-hermes-agent-on-luci-384-commits-9a8f7d Branch: cr/mc-4735-update-hermes-agent-on-luci-384-commits-9a8f7d tmux: cr-MC-4735 Expected check-in: 2026-06-05T10:20:02.012958+00:00

[control-room-recover] MC-4735: recovered DONE from a finished direct worker whose tmux was gone. Ticket → in_review. Summary: MC-4735 — Hermes updated (b1e399de→3da44dbda, v0.15.1) + iris config v24→v26 + recovered the wedged iris gateway (adopted Hermes-native user unit, disabled conflicting MC-4220 system unit, ported hardening via kernel-enforced drop-in). Council (Codex+Opus+Gemini+GLM) unanimous on approach. Smoke green: gateways running, telegram connected, child MCP up, NRestarts=0, MemoryMax enforced. Committed+p [control-room-merge-required] ⚠️ NOT SHIPPED: the work is committed on `cr/mc-4735-update-hermes-agent-on-luci-384-commits-9a8f7d` but is NOT on master. This ticket is review/merge-ready, not done. Land it safely with: python3 mission-control/scripts/control_room_pickup.py --merge-recovered (or `git -C /home/lucienne/workspace fetch && git -C /home/lucienne/workspace merge --ff-only cr/mc-4735-update-hermes-agent-on-luci-384-commits-9a8f7d` from a master worktree, then push). `done` is blocked until the branch is an ancestor of origin/master.

Controller gate: not closing MC-4735 yet. Live verification shows Hermes is running v0.15.1 and both hermes-gateway + hermes-gateway-iris are active, but `hermes --version` still reports update available (78 commits behind) and ~/.hermes/hermes-agent is behind origin/main. I routed the remaining technical update/verification work through Hermes Kanban mc-internal card t_ff3a8423 assigned to claudebuilder (run #72, pid 4043833). No Elmar input needed unless the worker finds a real approval/credential blocker.

Controller recovery: internal Kanban task t_ff3a8423 was blocked after claudebuilder exited twice without a valid kanban_complete/kanban_block handoff. I reassigned it to codexbuilder, unblocked it, and dispatched it through Hermes Kanban mc-internal. MC remains the ledger only; no MC ticket runtime/pickup was launched.

Controller gate complete (visibility-only): internal Kanban task t_ff3a8423 finished and I independently verified the Hermes Agent update state without using MC pickup/runtime launch. Evidence: - `hermes --version` and `hermes --profile iris --version` both report Hermes Agent v0.16.0 (2026.6.5), upstream ef7e5168, Up to date. - `hermes update --check` => Already up to date. - /home/lucienne/.hermes/hermes-agent HEAD is ef7e5168b52e4d3b38028fb960b48af88b149154 on main tracking origin/main, ahead/behind 0/0. - hermes-gateway and hermes-gateway-iris are active/running with NRestarts=0; iris MemoryMax=1153433600. - Iris hardening drop-in still contains MemoryHigh=750M, MemoryMax=1100M, NoNewPrivileges=true, PrivateTmp=true, KillMode=mixed, TimeoutStopSec=240. No MC runtime-start/send/claim/pickup was used. No new cron jobs were created.