Fix flaky test_multi_timer_validation on Windows CI

Increase timer tolerance from 150ms to 500ms to account for CI
environments under heavy load, especially Windows during wheel
building where timing can be significantly less precise.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: dkropachev

tests/unit/io/utils.py

@@ -128,8 +128,10 @@ def submit_and_wait_for_completion(unit_test, create_timer, start, end, incremen
         time.sleep(.1)
 
     # ensure they are all called back in a timely fashion
+    # Use a generous tolerance (500ms) to account for CI environments under heavy load,
+    # especially Windows during wheel building where timing can be significantly less precise
     for callback in completed_callbacks:
-        assert callback.expected_wait == pytest.approx(callback.get_wait_time(), abs=.15)
+        assert callback.expected_wait == pytest.approx(callback.get_wait_time(), abs=.5)
 
 
 def noop_if_monkey_patched(f):