diff --git a/leenkx/blender/lnx/handlers.py b/leenkx/blender/lnx/handlers.py
index ef47589..5d62a25 100644
--- a/leenkx/blender/lnx/handlers.py
+++ b/leenkx/blender/lnx/handlers.py
@@ -2,7 +2,10 @@ import importlib
 import os
 import queue
 import sys
+import threading
+import time
 import types
+from typing import Dict, Tuple, Callable, Set
 
 import bpy
 from bpy.app.handlers import persistent
@@ -30,6 +33,10 @@ if lnx.is_reload(__name__):
 else:
     lnx.enable_reload(__name__)
 
+# Module-level storage for active threads (eliminates re-queuing overhead)
+_active_threads: Dict[threading.Thread, Callable] = {}
+_last_poll_time = 0.0
+_consecutive_empty_polls = 0
 
 @persistent
 def on_depsgraph_update_post(self):
@@ -135,35 +142,113 @@ def always() -> float:
 
 
 def poll_threads() -> float:
-    """Polls the thread callback queue and if a thread has finished, it
-    is joined with the main thread and the corresponding callback is
-    executed in the main thread.
     """
+    Improved thread polling with:
+    - No re-queuing overhead
+    - Batch processing of completed threads
+    - Adaptive timing based on activity
+    - Better memory management
+    - Simplified logic flow
+    """
+    global _last_poll_time, _consecutive_empty_polls
+    current_time = time.time()
+    
+    # Process all new threads from queue at once (batch processing)
+    new_threads_added = 0
     try:
-        thread, callback = make.thread_callback_queue.get(block=False)
+        while True:
+            thread, callback = make.thread_callback_queue.get(block=False)
+            _active_threads[thread] = callback
+            new_threads_added += 1
     except queue.Empty:
+        pass
+    
+    # Early return if no active threads
+    if not _active_threads:
+        _consecutive_empty_polls += 1
+        # Adaptive timing: longer intervals when consistently empty
+        if _consecutive_empty_polls > 10:
+            return 0.5  # Back off when no activity
         return 0.25
-    if thread.is_alive():
-        try:
-            make.thread_callback_queue.put((thread, callback), block=False)
-        except queue.Full:
-            return 0.5
-        return 0.1 
+    
+    # Reset empty poll counter when we have active threads
+    _consecutive_empty_polls = 0
+    
+    # Find completed threads (single pass, no re-queuing)
+    completed_threads = []
+    for thread in list(_active_threads.keys()):
+        if not thread.is_alive():
+            completed_threads.append(thread)
+    
+    # Batch process all completed threads
+    if completed_threads:
+        _process_completed_threads(completed_threads)
+    
+    # Adaptive timing based on activity level
+    active_count = len(_active_threads)
+    if active_count == 0:
+        return 0.25
+    elif active_count <= 3:
+        return 0.05  # Medium frequency for low activity
     else:
+        return 0.01  # High frequency for high activity
+
+def _process_completed_threads(completed_threads: list) -> None:
+    """Process a batch of completed threads with robust error handling."""
+    for thread in completed_threads:
+        callback = _active_threads.pop(thread)  # Remove from tracking
+        
         try:
-            thread.join()
+            thread.join()  # Should be instant since thread is dead
             callback()
         except Exception as e:
-            # If there is an exception, we can no longer return the time to
-            # the next call to this polling function, so to keep it running
-            # we re-register it and then raise the original exception.
-            try:
-                bpy.app.timers.unregister(poll_threads)
-            except ValueError:
-                pass
-            bpy.app.timers.register(poll_threads, first_interval=0.01, persistent=True)
-        # Quickly check if another thread has finished
-        return 0.01
+            # Robust error recovery
+            _handle_callback_error(e)
+            continue  # Continue processing other threads
+        
+        # Explicit cleanup for better memory management
+        del thread, callback
+
+def _handle_callback_error(exception: Exception) -> None:
+    """Centralized error handling with better recovery."""
+    try:
+        # Try to unregister existing timer
+        bpy.app.timers.unregister(poll_threads)
+    except ValueError:
+        pass  # Timer wasn't registered, that's fine
+    
+    # Re-register timer with slightly longer interval for stability
+    bpy.app.timers.register(poll_threads, first_interval=0.1, persistent=True)
+    
+    # Re-raise the original exception after ensuring timer continuity
+    raise exception
+
+def cleanup_polling_system() -> None:
+    """Optional cleanup function for proper shutdown."""
+    global _active_threads, _consecutive_empty_polls
+    
+    # Wait for remaining threads to complete (with timeout)
+    for thread in list(_active_threads.keys()):
+        if thread.is_alive():
+            thread.join(timeout=1.0)  # 1 second timeout
+    
+    # Clear tracking structures
+    _active_threads.clear()
+    _consecutive_empty_polls = 0
+    
+    # Unregister timer
+    try:
+        bpy.app.timers.unregister(poll_threads)
+    except ValueError:
+        pass
+
+def get_polling_stats() -> dict:
+    """Get statistics about the polling system for monitoring."""
+    return {
+        'active_threads': len(_active_threads),
+        'consecutive_empty_polls': _consecutive_empty_polls,
+        'thread_ids': [t.ident for t in _active_threads.keys()]
+    }
 
 
 loaded_py_libraries: dict[str, types.ModuleType] = {}