Fix watchdog relaunch-loop: guard before relaunch, main-process PID filter

The relaunch-loop guard only protected the kill step, not the relaunch.
When CDP was unreachable on a slow-starting browser, the watchdog would
invoke netvm-chrome.sh (which kills the existing browser) before checking
if it was recently launched — piling up 5 chromiums on opm.

Now the <120s check runs before the relaunch and skips the entire cycle.
Also fixed the PID check to match only the main browser process
(--remote-debugging-port, excluding --type= renderer/gpu children).

Session: sidechat/chromebox-ops

bin/chromebox-watchdog.sh

@@ -75,12 +75,21 @@ if healthy; then
   exit 0
 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).
+# Relaunch-loop guard (2026-10-04): if the main browser for this profile
+# launched <2 min ago it's probably still starting up (CDP not yet bound).
+# Relaunching now would kill a healthy-but-slow cold start via netvm-chrome.sh
+# and reset the startup clock every cycle (observed: opm piled up 5 chromiums
+# because the guard only protected the kill step, not the relaunch). Skip the
+# entire cycle instead.
+# NOTE: match only the main browser process (--remote-debugging-port present,
+# no --type= flag). Renderer/gpu children (--type=renderer etc.) start later
+# than the main process and must not satisfy this check.
 recent_pid=""
-for _pid in $(pgrep -f "chromium.*profiles/${PROFILE}/" 2>/dev/null); do
+for _pid in $(pgrep -f "chromium.*--remote-debugging-port=${CDP_PORT}([[:space:]]|$)" 2>/dev/null); do
+  # Skip child processes (renderer, gpu, etc.) — only the main browser counts
+  if ps -o args= -p "$_pid" 2>/dev/null | grep -q -- "--type="; then
+    continue
+  fi
   _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
@@ -89,9 +98,10 @@ for _pid in $(pgrep -f "chromium.*profiles/${PROFILE}/" 2>/dev/null); do
   fi
 done
 if [ -n "$recent_pid" ]; then
-  log "browser launched recently (pid $recent_pid), skipping kill (probably still starting)"
+  log "browser launched recently (pid $recent_pid), skipping relaunch (probably still starting)"
   exit 0
 fi
+log "unhealthy ($HEALTH_FAIL_REASON), relaunching chromebox"
 # 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