chromebox-watchdog: fix kill loop on slow cold starts

Two fixes: (1) retry health check 4x with 15s gaps after relaunch instead of single 25s check; (2) skip kill if browser launched <2min ago (probably still starting). Prevents watchdog from killing a working-but-slow browser.

Session: sidechat/chromebox-ops

bin/chromebox-watchdog.sh

@@ -72,6 +72,22 @@ if healthy; then
 fi
 
 log "unhealthy ($HEALTH_FAIL_REASON), relaunching chromebox"
+# Kill-loop guard (2026-10-04): if a chromium for this profile launched <2 min
+# ago it's probably still loading the page — killing it just restarts the
+# loop (observed: 646 killed twice while the page was still loading).
+recent_pid=""
+for _pid in $(pgrep -f "chromium.*profiles/${PROFILE}/" 2>/dev/null); do
+  _start=$(date -d "$(ps -o lstart= -p "$_pid" 2>/dev/null)" +%s 2>/dev/null || echo 0)
+  _now=$(date +%s)
+  if [ $(( _now - _start )) -lt 120 ] && [ "$_start" -gt 0 ]; then
+    recent_pid="$_pid"
+    break
+  fi
+done
+if [ -n "$recent_pid" ]; then
+  log "browser launched recently (pid $recent_pid), skipping kill (probably still starting)"
+  exit 0
+fi
 # bracket trick so pkill never matches its own command line
 pat="profiles/${PROFILE:0:${#PROFILE}-1}[${PROFILE: -1}]/"
 pkill -f "chromium.*$pat" 2>/dev/null || true
@@ -93,10 +109,21 @@ sleep 3
 systemd-run --user --scope --unit="netvm-chrome-${PROFILE}-$(date +%s)" \
   "$NETVM_BIN/netvm-chrome.sh" --headless --cdp-port "$CDP_PORT" "$PROFILE" "https://muse.ai" \
   >>"$CHROME_LOG" 2>&1 < /dev/null &
-sleep 25
+# Retry with backoff (2026-10-04): cold starts (fresh egress IP, Cloudflare
+# handshake) can take >25s for the page title to appear. A single check after
+# 25s kills working-but-slow browsers. Try up to 4 times, 15s apart (~60s
+# window), logging each attempt. Only declare FAILED if all attempts fail.
+_relaunch_ok=0
+for _attempt in 1 2 3 4; do
+  sleep 15
   if healthy; then
-  log "relaunch OK"
+    log "relaunch OK (attempt $_attempt)"
-else
+    _relaunch_ok=1
+    break
+  fi
+  log "relaunch attempt $_attempt not healthy yet ($HEALTH_FAIL_REASON)"
+done
+if [ "$_relaunch_ok" -ne 1 ]; then
   log "relaunch FAILED ($HEALTH_FAIL_REASON) \u2014 needs operator attention"
   exit 1
 fi