This release introduces significant performance optimizations for high-concurrency environments and large data operations, including faster multipart uploads and more efficient IAM checks. New administrative controls provide greater flexibility for managing cluster health and storage pool decommissioning, while key bug fixes enhance data management reliability and system responsiveness.

### New Features
*   **Customize Health Check Interval**: Configure the internode health check interval for faster detection of unresponsive nodes and reduced cluster failover times. (#851)
*   **Accelerate Pool Decommissioning**: Optionally skip the data cleanup phase on a source pool to significantly accelerate storage pool decommissioning and data migration workflows. Note: Data will persist on the decommissioned pool and requires manual cleanup to reclaim space. (#871)
*   **Enhanced Replication Metrics**: Gain deeper visibility into data replication with new metrics for 'in progress' tasks and detailed target information per node, enabling more effective monitoring. (#845)
*   **Proactive License Management**: The MinIO license ID is now included in cluster registration data, enabling the SUBNET service to provide timely warnings for license expirations. (#875)

### Improvements
*   **Expanded Elasticsearch Compatibility**: Upgraded event notification system to support Elasticsearch server versions 7, 8, and 9 for broader integration options. (#820)
*   **Improved Drive Health Detection**: Health checks now monitor read and write operations independently, ensuring more accurate and timely detection of partially failing drives. (#848)
*   **Robust LDAP Configuration Validation**: Strengthened validation for LDAP settings to prevent misconfigurations and ensure more predictable authentication behavior. (#776)

### Performance
*   **Optimized IAM Verification**: Reduced CPU utilization and improved server responsiveness in large-scale, high-concurrency environments by optimizing IAM key checks. (#863)
*   **Faster Multipart Uploads**: Increased multipart upload speed by reducing network calls required for managing temporary part metadata. (#889)
*   **Default Multipart Caching**: Improved multipart upload performance by enabling the multipart cache (`mpCache`) by default for standard deployments. (#866)
*   **Accelerated Empty Bucket Deletion**: Optimized the check for bucket contents to significantly speed up the deletion of empty buckets, especially in large clusters. (#878)
*   **Efficient Trash Handling**: Implemented rotating prefixes for trashed items to prevent performance degradation during large-scale data deletion operations. (#877)
*   **Reduced Disk Scan Overhead**: Eliminated unnecessary concurrent disk scans to improve system responsiveness, particularly in deployments with numerous storage sets. (#887)

### Bug Fixes
*   **Data Healing and Read Operations**: Modified the locking strategy during data healing to prioritize read operations, minimizing impact on concurrent data access. Healing may take longer for actively read objects. (#879)
*   **Multipart Uploads in Stretched Clusters**: Resolved an issue causing '503 Service Unavailable' errors during large multipart uploads in high-latency, stretched cluster environments. (#861)
*   **Encrypted Metadata Access**: Fixed an incompatibility with older metadata formats to ensure reliable access to all encrypted objects. (#882, #885)
*   **Site Replication Timeouts**: Resolved premature timeouts during site replication to improve success rates for large datasets or across high-latency networks. (#874)
*   **Lifecycle Management Processing**: Ensured object lifecycle management (ILM) expiration rules continue processing without error when a target object has already been deleted. (#855)
*   **Bucket Quota Enforcement**: Corrected an issue where bucket quotas set with older software versions were not enforced after an upgrade. (#870)
*   **Non-Versioned Object Deletion**: Fixed a '503 Service Unavailable' error that could occur when deleting non-versioned objects while a storage node is offline. (#854)
*   **Server Crash on Full Disk**: Prevented a server crash that could occur when moving files to the trash on a nearly full storage drive. (#896)
*   **Resource Leak from Background Tasks**: Prevented a resource leak by ensuring background tasks are properly cleaned up after internal errors. (#892)
*   **Restored Admin APIs**: Re-enabled key administrative API endpoints, including `/minio/admin/v4/object-summary` (#859) and OIDC endpoints `/idp/openid/list-access-keys-bulk` and `/info-access-key` (#857).
*   **Batch Catalog Operation Timeouts**: Fixed a regression that caused certain batch catalog operations to fail due to premature timeouts. (#894)
*   **MinIO Client (mc) Stability**: Resolved a client crash that occurred when rendering diagnostic data in a table format. (#864)
*   **Excessive Audit Logging**: Reduced unnecessary audit logging during internal tiered object decommissioning, resulting in cleaner and more focused audit trails. (#880)
*   **Elasticsearch Index Creation**: Improved error handling to ensure failures during Elasticsearch index creation are correctly reported. (#888)

### Security
*   **Hardened Kubernetes Authentication**: Enhanced security by defaulting the expected audience for Kubernetes service account tokens to `sts.min.io`. To maintain compatibility with tokens lacking a specific audience, administrators can set the `kubernetes token_audience` to an empty string, though this is deprecated. (#674)