MC-4853

[visibility-only controller] Critical scheduler timeout routed to internal Hermes Kanban card t_512e01ae (assignee=codexbuilder), running in isolated worktree `/home/lucienne/workspace/state/kanban-worktrees/mc-4853-canary-board-regression-timeout` (verified worker pid 2495321 cwd matches). Source: orchestrator_inbox 150271 / task_run 395618, canary-board-regression killed after 360s. Worker is to find/fix the deterministic canary timeout or block with exact evidence. MC remains visibility-only; no MC runtime/pickup endpoints were used.

[visibility-only controller] Follow-up gate for internal card t_512e01ae: it completed read-only and identified the canary timeout root cause (scheduler self-heal consumed the 360s budget after a transient fixture-ticket POST/connection-reset failure), but it did not land a hardening fix. Routed the technical follow-up to mc-internal Kanban card t_74efbd9e (assignee=codexbuilder) in isolated workspace `/home/lucienne/workspace/state/kanban-worktrees/mc-4853-canary-board-regression-hardening`; verified worker pid 2509622 cwd matches. Also reclaimed/archived an accidental duplicate created on the default Kanban board before continuing on mc-internal. MC remains visibility-only; no MC runtime/pickup endpoints were used.

[visibility-only controller] Internal Kanban review for MC-4853 rejected the prior t_74efbd9e handoff and requeued the same card to codexbuilder. Evidence: `python3 -m pytest tests/test_mc_ticket_regressions.py tests/test_scheduler_regressions.py::SchedulerRuntimeFailureTests::test_canary_board_regression_disables_scheduler_self_heal -q` failed in the isolated top-level worktree because the regression imported `/home/lucienne/workspace/mc_ticket.py` from the live checkout rather than the worktree copy, so the claimed ConnectionResetError guard was not actually proven. Worker has been instructed to fix the test/import isolation and rerun both top-level and Mission Control canary tests from clean/isolated checkouts. MC remains visibility-only; no MC runtime/pickup endpoints were used.

[visibility-only controller] MC-4853 gate PASS and closed. Evidence: internal Kanban t_74efbd9e completed after independent review; top-level luci-workspace commit 5a3f0309 and mission-control commit 8d5ff2d were pushed to origin/master and live checkouts advanced to those SHAs. Verified: top-level unittest/pytest regressions (2 passed), mission-control canary unit tests (2 passed), and live `python3 /home/lucienne/workspace/mission-control/tests/canary/run_board_battery.py` returned BATTERY pass=12 fail=0 consoleErrors=0. Preserved pre-existing overlapping live top-level diff at `/home/lucienne/workspace/ticket-attachments/MC-4853/live-top-level-overlap-before-deploy-20260608T2215Z.patch` sha256=77965b548aa8349d78c44e4b2d540d52dc5fab2d8e020e4e341e9e1c88adc864 before deploying selected files. `graphify update .` completed in mission-control; top-level graphify update was attempted but exceeded the 600s tool timeout after tests passed. No MC runtime/pickup endpoints were used.