CloudWatch alarm metric evaluation varies from AWS behavior

CloudWatch alarm triggering values in LocalStack can differ from AWS due to timing differences in metric ingestion and evaluation windows. Alarm values may vary (e.g., 21 vs expected average 21.5), causing test failures at approximately 14% rate.

Reduce alarm evaluation period to 10 seconds or increase alarm threshold to account for LocalStack's metric evaluation timing differences. Adjust test expectations to accommodate variation between LocalStack and AWS behavior.