Worked on the wazuh/wazuh-indexer repository to improve replication reliability during rolling restarts by addressing a bug that caused segment replication failures. Developed a targeted fix in Java that distinguishes between normal replication and recovery scenarios, ensuring that stale checkpoints are handled appropriately based on the shard’s state. This approach reduced allocation failures and improved cluster uptime and data integrity. Enhanced the codebase with comprehensive unit tests and updated mocks to verify correct behavior for both active and recovering shards. Documented the changes in the project’s CHANGELOG, demonstrating a methodical approach to backend development, mocking, and unit testing practices.