docs(handoff): add Ferrox ADF health-check handoff

Co-Authored-By: Claude <noreply@anthropic.com>

Author: claude

.docs/handoff-ferrox-adf-health.md

@@ -0,0 +1,35 @@
+# Incomplete Handoff: ADF Meta-Coordinator Health Checks
+
+## Agent
+Ferrox (Rust Engineer)
+
+## Done
+- Identified root cause: `scripts/adf-meta-coordinator-health.sh` used `sudo journalctl`, which failed silently in the containerized/no-new-privileges environment, making the check blind.
+- Added `journal_adf()` helper that probes `sudo -n` first and falls back to non-sudo `journalctl`.
+- Committed fix as `36196f049` (after merging `main`):
+  - `scripts/adf-meta-coordinator-health.sh`
+- Pushed to `origin` and `gitea` (both remotes converged; `git diff origin/main gitea/main --stat` is empty).
+- Ran the fixed script and created health-alert issues:
+  - #3055: [ADF] Tick-stall detected: 123 in 4h
+  - #3056: [ADF] 46 agent failures in 4h
+- Closed related meta-issues as resolved:
+  - #3051, #3052, #3054
+
+## Current System State (last 4h)
+- 122–124 `reconcile_tick SLOW` stalls
+- 46 non-success agent exits (mostly `model_error` / `server error`, some `timeout` in `security-sentinel`)
+- 0 `max_cpu_seconds` timeouts
+- 12–15 worktrees present, ~1 GB disk usage (under 10 GB threshold)
+- `adf` binary version correct: `adf 1.20.2`
+- Gitea API healthy (HTTP 200)
+
+## Remaining Work
+1. **Investigate and remediate the actual health findings**: the tick stalls and agent failures are real production issues, not artifacts of the broken script. Root-cause analysis needed for:
+   - Why `reconcile_tick` is slow 122+ times in 4h.
+   - Why agents (especially `product-development`, `implementation-swarm`, `security-sentinel`, `terraphim-agents-developer`, `test-guardian`) are exiting non-success.
+2. **Clean up issue bodies**: #3055 and #3056 were created by the pre-fix script and contain literal `\n` sequences instead of real newlines. Low priority; can be edited via Gitea API for readability.
+
+## Next-Agent Starting Position
+- Branch/commit: `main` @ `36196f049` on this worktree.
+- Start with the health-alert issues #3055 and #3056; inspect the attached log excerpts and `journalctl -u adf-orchestrator` for the relevant window.
+- Consider involving the orchestrator maintainer for the `reconcile_tick SLOW` root cause.