diff --git a/docs/LOGBOOK.md b/docs/LOGBOOK.md
index 6660bbfd..38fd1c30 100644
--- a/docs/LOGBOOK.md
+++ b/docs/LOGBOOK.md
@@ -11,6 +11,7 @@
 - StockPlatform public edge / API 已恢復：`https://stock.wooo.work/healthz=200`、`/api/healthz=200`；但 freshness / ingestion 仍 `status=not_configured`，blocker `postgres_not_ready`。
 
 **source 修正**：
+- `awoooi-startup-110.sh` 新增 bounded Docker start、dockerd journal corrupt metadata ID 抽取、container metadata dir quarantine 與 TERM main process 後重啟 Docker 的受控恢復路徑，避免下次 dockerd 因 JSON metadata 損壞卡死 startup。
 - `full-stack-cold-start-check.sh` 新增 registry 外部 `/v2/` gate 與 K3s `IMAGE_PULL_BLOCKED` / `REGISTRY_PULL_REFUSED_EVENTS` blocker。
 - `full-stack-recovery-scorecard.sh` 新增 `CORE_REGISTRY_HTTPS_CODE`、`CORE_REGISTRY_HTTP_CODE`、`CORE_REGISTRY_READY`、`CORE_REGISTRY_BLOCKER`，registry `:5000` 未就緒時不允許 `CORE_READY`。
 
diff --git a/scripts/reboot-recovery/awoooi-startup-110.sh b/scripts/reboot-recovery/awoooi-startup-110.sh
index ae8ab647..bc6a9a78 100644
--- a/scripts/reboot-recovery/awoooi-startup-110.sh
+++ b/scripts/reboot-recovery/awoooi-startup-110.sh
@@ -6,15 +6,99 @@
 #
 # 已知問題處理:
 # - Docker BoltDB 損壞 (network/files/local-kv.db, volumes/metadata.db)
+# - Docker container metadata JSON 損壞，dockerd 卡在 startup / loading containers
 # - 舊容器使用已不存在的 Docker network (需要 docker rm -f 全部)
 # - Harbor nginx 依賴 harbor-log (需要等 harbor-log healthy 後才 compose up)
 
 set -uo pipefail
 LOG="/var/log/awoooi-startup-110.log"
+DOCKER_START_TIMEOUT_SECONDS="${AWOOOI_DOCKER_START_TIMEOUT_SECONDS:-45}"
+DOCKER_ACTIVE_WAIT_ATTEMPTS="${AWOOOI_DOCKER_ACTIVE_WAIT_ATTEMPTS:-18}"
+DOCKER_ACTIVE_WAIT_SLEEP_SECONDS="${AWOOOI_DOCKER_ACTIVE_WAIT_SLEEP_SECONDS:-5}"
 exec > >(tee -a "$LOG") 2>&1
 
 log() { echo "[$(date '+%Y-%m-%d %H:%M:%S')] $*"; }
 
+run_bounded() {
+    local seconds="$1"
+    shift
+    if command -v timeout >/dev/null 2>&1; then
+        timeout "$seconds" "$@"
+    else
+        "$@"
+    fi
+}
+
+start_docker_bounded() {
+    log "bounded start docker.service/docker.socket timeout=${DOCKER_START_TIMEOUT_SECONDS}s"
+    run_bounded "$DOCKER_START_TIMEOUT_SECONDS" systemctl start docker.socket docker.service || true
+}
+
+wait_docker_active() {
+    local i state
+    for i in $(seq 1 "$DOCKER_ACTIVE_WAIT_ATTEMPTS"); do
+        state=$(systemctl is-active docker 2>/dev/null || true)
+        log "Docker wait ${i}/${DOCKER_ACTIVE_WAIT_ATTEMPTS}: ${state:-unknown}"
+        [ "$state" = "active" ] && return 0
+        sleep "$DOCKER_ACTIVE_WAIT_SLEEP_SECONDS"
+    done
+    return 1
+}
+
+stop_docker_bounded_for_metadata_recovery() {
+    local state
+    run_bounded 25 systemctl stop docker.service docker.socket || true
+    state=$(systemctl is-active docker 2>/dev/null || true)
+    if [ "$state" = "active" ] || [ "$state" = "activating" ] || [ "$state" = "deactivating" ]; then
+        log "Docker stop still ${state}; TERM main process for metadata quarantine recovery"
+        systemctl kill --kill-who=main --signal=TERM docker.service || true
+        sleep 4
+        run_bounded 20 systemctl stop docker.service docker.socket || true
+    fi
+}
+
+DOCKER_CORRUPT_METADATA_QUARANTINED_COUNT=0
+quarantine_corrupt_docker_container_metadata() {
+    local ids qdir id src count
+    DOCKER_CORRUPT_METADATA_QUARANTINED_COUNT=0
+    ids=$(journalctl -u docker.service --since "-30 min" --no-pager 2>/dev/null \
+        | awk '/failed to load container|invalid character|json: cannot unmarshal|unexpected EOF/ {
+            line=$0
+            while (match(line, /[0-9a-f]{64}/)) {
+                print substr(line, RSTART, RLENGTH)
+                line=substr(line, RSTART + RLENGTH)
+            }
+        }' \
+        | sort -u)
+
+    if [ -z "$ids" ]; then
+        log "Docker corrupt container metadata: journal 未找到可 quarantine 的 container id"
+        return 0
+    fi
+
+    qdir="/var/lib/docker/containers/.awoooi-corrupt-metadata-quarantine-$(date +%Y%m%d-%H%M%S)"
+    mkdir -p "$qdir"
+    count=0
+    for id in $ids; do
+        case "$id" in
+            *[!0-9a-f]*|"")
+                log "跳過不合法 container id: $id"
+                continue
+                ;;
+        esac
+        src="/var/lib/docker/containers/$id"
+        if [ -d "$src" ]; then
+            mv "$src" "$qdir/$id"
+            count=$((count + 1))
+            log "quarantine corrupt Docker container metadata: $id"
+        else
+            log "corrupt metadata candidate already absent: $id"
+        fi
+    done
+    DOCKER_CORRUPT_METADATA_QUARANTINED_COUNT="$count"
+    log "Docker corrupt metadata quarantine dir=$qdir count=$count"
+}
+
 log "=== 192.168.0.110 啟動序列開始 ==="
 
 # ──────────────────────────────────────────────
@@ -24,8 +108,18 @@ log "[1/5] 檢查 Docker..."
 
 if ! systemctl is-active docker >/dev/null 2>&1; then
     log "Docker 未啟動，嘗試啟動..."
-    systemctl start docker || true
-    sleep 8
+    start_docker_bounded
+    wait_docker_active || true
+fi
+
+if ! systemctl is-active docker >/dev/null 2>&1; then
+    log "Docker bounded start 未 active，檢查 corrupt container metadata..."
+    quarantine_corrupt_docker_container_metadata
+    if [ "$DOCKER_CORRUPT_METADATA_QUARANTINED_COUNT" -gt 0 ]; then
+        stop_docker_bounded_for_metadata_recovery
+        start_docker_bounded
+        wait_docker_active || true
+    fi
 fi
 
 if ! systemctl is-active docker >/dev/null 2>&1; then
@@ -47,8 +141,8 @@ if ! systemctl is-active docker >/dev/null 2>&1; then
     # 清除 buildkit 快取（也可能損壞）
     find /var/lib/docker/buildkit -name "*.db" -delete 2>/dev/null || true
 
-    systemctl start docker
-    sleep 8
+    start_docker_bounded
+    wait_docker_active || true
     systemctl is-active docker && log "✅ Docker 修復成功" || { log "❌ Docker 修復失敗"; exit 1; }
 else
     log "✅ Docker 已 active"
diff --git a/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py b/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py
index 73bd71c4..f668b55d 100644
--- a/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py
+++ b/scripts/reboot-recovery/tests/test_cold_start_monitor_bounded_probes.py
@@ -5,6 +5,7 @@ from pathlib import Path
 
 ROOT = Path(__file__).resolve().parents[3]
 COLD_START_CHECK = ROOT / "scripts" / "reboot-recovery" / "full-stack-cold-start-check.sh"
+STARTUP_110 = ROOT / "scripts" / "reboot-recovery" / "awoooi-startup-110.sh"
 RECOVERY_SCORECARD = (
     ROOT / "scripts" / "reboot-recovery" / "full-stack-recovery-scorecard.sh"
 )
@@ -57,6 +58,18 @@ def test_full_stack_cold_start_check_tracks_registry_and_image_pull_blockers() -
     assert "K3s registry pull refused by 110:5000" in text
 
 
+def test_startup_110_quarantines_corrupt_docker_container_metadata() -> None:
+    text = STARTUP_110.read_text(encoding="utf-8")
+
+    assert "AWOOOI_DOCKER_START_TIMEOUT_SECONDS" in text
+    assert "quarantine_corrupt_docker_container_metadata()" in text
+    assert "DOCKER_CORRUPT_METADATA_QUARANTINED_COUNT" in text
+    assert "failed to load container|invalid character|json: cannot unmarshal|unexpected EOF" in text
+    assert 'mv "$src" "$qdir/$id"' in text
+    assert "systemctl kill --kill-who=main --signal=TERM docker.service" in text
+    assert 'run_bounded "$DOCKER_START_TIMEOUT_SECONDS" systemctl start docker.socket docker.service' in text
+
+
 def test_cold_start_deploy_parity_verifier_bounds_ssh_readback() -> None:
     text = VERIFY_DEPLOY.read_text(encoding="utf-8")