fix(dashboard): harden generation against common failures

SAC_LSTM_Bot/tools/watch_dashboard.sh

@@ -1,10 +1,23 @@
 #!/bin/sh
 # Keep docs/campaign_dashboard.svg fresh: regenerate every 60 s.
-# Errors go to stderr and never exit the loop silently.
-dir=$(dirname "$0")
+# Failures are appended to tools/plot_progress.log (by this wrapper and by
+# plot_progress.py itself) - check that file when the dashboard looks stale.
+# The startup banner records WHICH tree this instance watches, so a copy
+# launched from a stale checkout (the post-reboot failure mode) is visible.
+set -u
+dir=$(cd "$(dirname "$0")" && pwd)
+log="$dir/plot_progress.log"
+
+if command -v flock >/dev/null 2>&1; then
+    exec 9>"$dir/.watch_dashboard.lock"
+    flock -n 9 || { echo "[watch_dashboard] another instance already running, exiting" >&2; exit 0; }
+fi
+
+echo "[watch_dashboard] $(date '+%F %T') started, watching root=$(cd "$dir/.." && pwd)" >> "$log" 2>/dev/null || true
 while :; do
     if ! python3 "$dir/plot_progress.py"; then
-        echo "[watch_dashboard] $(date '+%F %T') regeneration failed (see error above)" >&2
+        echo "[watch_dashboard] $(date '+%F %T') regeneration failed (see $log)" >&2
+        echo "[watch_dashboard] $(date '+%F %T') regeneration failed" >> "$log" 2>/dev/null || true
     fi
     sleep 60
 done