Resilience pass: bulletproof recovery + FWS / fault polling + EG4 discovery refresh

LVX6048/bin/lvx-resolve-links

@@ -89,25 +89,44 @@ async def main() -> int:
             break
         await asyncio.sleep(0.5)
 
-    if not sn_to_path:
-        print("no LVX6048 devices found on /dev/hidraw*", file=sys.stderr)
-        return 1
-
-    missing = []
+    # Always-best-effort policy: exit 0 even if no expected serials were
+    # found, so a transient inverter blip (e.g. the unit being mid-cold-start
+    # at sunrise) doesn't permanently latch dependent services into a failed
+    # state via systemd `Requires=`. The periodic timer + powermon's own
+    # Restart=always handles convergence once the inverter recovers.
+    changed = False
     for sn, link in LINK_FOR_SERIAL.items():
         if sn in sn_to_path:
-            _relink(link, sn_to_path[sn])
-            print(f"symlink {link} -> {os.path.basename(sn_to_path[sn])}")
+            target = sn_to_path[sn]
+            current = os.readlink(link) if os.path.islink(link) else None
+            current_full = os.path.join(os.path.dirname(link), current) if current else None
+            if current_full != target:
+                _relink(link, target)
+                print(f"symlink {link} -> {os.path.basename(target)}")
+                changed = True
+            else:
+                print(f"symlink {link} -> {os.path.basename(target)} (unchanged)")
         else:
-            missing.append((link, sn))
-            try:
-                if os.path.islink(link):
+            if os.path.islink(link):
+                try:
                     os.unlink(link)
-            except FileNotFoundError:
-                pass
+                    changed = True
+                except FileNotFoundError:
+                    pass
             print(f"WARNING: {link} serial {sn} not found on any /dev/hidraw*")
 
-    return 0 if not missing else 2
+    # If symlinks actually moved (e.g. hidraw indices flipped post-power-cycle),
+    # bounce powermon so its open hidraw fds re-bind to the right physical unit.
+    # Idempotent runs (no symlink change) leave powermon alone.
+    if changed:
+        print("symlinks changed — restarting powermon services")
+        import subprocess
+        subprocess.Popen(
+            ["/bin/systemctl", "--no-block", "restart",
+             "powermon.service", "powermon2.service"]
+        )
+
+    return 0
 
 
 if __name__ == "__main__":