This release introduces significant enhancements to data management, scalability, and performance. You can now apply Object Lock to existing buckets for improved data immutability and use flexible string-based versions for KES keys. Performance is enhanced through optimized write operations and faster bulk deletes, while new tiered fanout calls improve scalability for large cluster communications.

### New Features

- **Enabled Object Lock on existing buckets**, allowing the application of WORM retention policies without data migration. Note: Retention settings must be configured on objects after enabling this feature at the bucket level. (#840)
- **Introduced string-based key versioning in the Key Encryption Service (KES)**, providing greater flexibility for custom key lifecycle management and integration with external versioning schemes. (#797)
- **Added a streaming API for efficient, persistent data exchange**, reducing connection overhead for workflows with frequent transactions compared to traditional REST calls. (#806)
- **Implemented tiered fanout calls for cluster-wide communication**, improving scalability and reducing the load on any single node during broadcast operations in large clusters. (#807)
- **Allowed single-drive erasure sets for development and testing**, simplifying setup with minimal hardware. This configuration provides no data redundancy and is unsafe for production environments. (#837)

### Improvements

- **Improved management of very large folders** by logging and halting scans on folders exceeding size thresholds, preventing performance degradation and providing administrators with actionable metrics. (#821)
- **Enhanced error logging for object part decryption failures** to include the specific object key, accelerating troubleshooting and issue resolution. (#822)
- **Updated the management console with UI refinements** and corrected permission-handling logic for a more intuitive and predictable user experience. (#831)
- **Reduced telemetry data for public trial licenses** to only essential registration information, minimizing data collection from trial clusters. (#847)

### Performance Improvements

- **Removed an internal component to eliminate a significant throughput bottleneck**, increasing overall system performance across various setups. (#838)
- **Optimized memory allocation for specific erasure coding configurations** (e.g., 3, 5, 6, or 7 data shards), resulting in faster and more consistent write operations. (#833)
- **Accelerated bulk delete operations** by reducing network overhead and latency for improved system responsiveness. (#828)
- **Improved system-wide responsiveness under load** by accurately accounting for delete operation processing time, preventing server overload during high-activity periods. (#835)

### Bug Fixes

- **Resolved memory alignment issues that caused `O_DIRECT` call failures**, ensuring consistent high-performance I/O for demanding applications. (#836)
- **Prevented replication tasks from stalling indefinitely** by implementing a 15-minute request timeout, ensuring more consistent data synchronization. (#816)
- **Averted resource exhaustion under high request loads** by proactively respecting operating system file descriptor limits. (#844)
- **Corrected an issue that improperly blocked data appends** using the S3 `PutObject` API with the `write-offset-bytes` parameter. (#825)
- **Ensured Trino-initiated deletes correctly purge all object versions** in versioned buckets when legacy Spark optimization is active, preventing orphaned data. (#817)
- **Eliminated unnecessary IAM data reloads during startup**, improving initialization speed and user access on clusters with many IAM configurations. (#824)
- **Improved health check accuracy in distributed deployments** to correctly report isolated nodes as unavailable, preventing misrouted traffic. (#815)
- **Allowed replication configuration updates to proceed** even when other disabled replication targets are temporarily offline. (#823)
- **Standardized all system timestamps to UTC** to ensure data consistency in logs and interfaces across different server timezones. (#799)
- **Prevented redundant drive scanning** in single-drive configurations to improve background task efficiency. (#834)
- **Corrected the initial delay and authentication for `callhome` telemetry reporting**, ensuring timelier and more accurate communication for all license types. (#841)
- **Suppressed non-critical "unknown mux handler" log errors** in single-node deployments for cleaner operational logs. (#843)

### Security Updates

- **Updated core software dependencies to their latest versions**, incorporating upstream security patches and functional enhancements. (#842)