This release introduces significant enhancements to monitoring, performance, and administrative control. Administrators gain deeper operational insight with new Prometheus metrics for background jobs and replication, while performance is boosted for multipart uploads and high-load IAM operations. New features like customizable health checks and accelerated pool decommissioning provide greater flexibility for managing critical workflows.

### New Features

-   Customize the health check interval in distributed deployments to accelerate the detection of unresponsive nodes and reduce failover times. (#851)
-   Accelerate storage pool decommissioning by optionally skipping the data cleanup phase, enabling faster data migration when immediate availability on the new pool is critical. (#871)
-   Gain deeper visibility into background jobs with new Prometheus metrics for cataloging, expiry, key rotation, and replication, enabling proactive monitoring of job status and resource utilization. (#862)
-   Enhance replication monitoring with new metrics for 'in-progress' tasks and detailed target information, providing a clearer view of ongoing replication health. (#845)
-   Enable proactive license management with the inclusion of the license ID in cluster registration, allowing the SUBNET service to provide timely expiration warnings. (#875)
-   Verify object integrity with an expanded range of checksum algorithms (including SHA1, SHA256, and CRC variants) using the `s3-check-md5` tool for more flexible data validation. (#883)

### Performance Improvements

-   Optimized internal IAM key checks to reduce processing delays and improve server responsiveness under high load. (#863)
-   Improved multipart upload performance by enabling the multipart upload cache (`mpCache`) by default for standard deployments. (#866)
-   Increased multipart upload speed by optimizing the management of temporary part files and reducing underlying network calls. (#889)
-   Accelerated the deletion of empty buckets in large-scale systems by using a more efficient check to determine bucket status. (#878)
-   Improved data access responsiveness by modifying the locking strategy during object healing to prioritize concurrent read operations. (#879)
-   Increased system responsiveness in large-scale deployments by optimizing the management of concurrent operations and eliminating redundant disk scans. (#887)
-   Improved performance during heavy data deletion by using rotating prefixes for trashed items, preventing slowdowns from an excessive accumulation of files in a single directory. (#877)

### Improvements

-   Ensure modern compatibility by extending event notification support to Elasticsearch server versions 7, 8, and 9. (#820)
-   Strengthened LDAP integration with improved configuration validation, preventing misconfigurations from being applied. (#776)
-   Improved the robustness of software installation and update processes, particularly for deployments requiring custom scripts. (#906)
-   Updated all internal software libraries to their latest versions, incorporating security patches and performance optimizations. (#913)
-   The MinIO Console has been updated to the latest version, incorporating recent UI/UX improvements and fixes. (#918)

### Security Updates

-   Enhanced security for Kubernetes integrations by defaulting the service account token audience to `sts.min.io`. A new configuration option is available for backward compatibility, but updating to the new default is recommended for improved security. (#674)

### Bug Fixes

-   **Replication & Lifecycle:**
    -   Increased the reliability of site replication by extending internal timeouts, reducing failures for large datasets or across high-latency networks. (#874)
    -   Corrected bucket replication metrics to include the `targetArn` label, enabling precise, per-target monitoring and filtering. (#911)
    -   Corrected replication proxy error metrics for object tagging, preventing false increments when the target object or bucket does not exist. (#890)
    -   Ensured new ILM tier configurations are reliably applied across all nodes in a distributed deployment. (#902)
    -   Made the object lifecycle management (ILM) expiration process more resilient by ignoring already-deleted objects, preventing unnecessary errors and log noise. (#855)
-   **API & SDK:**
    -   Improved AWS SDK compatibility by consolidating throttling error codes to `SlowDown`, ensuring more reliable client-side retry behavior under load. (#903)
    -   Restored the `/minio/admin/v4/object-summary` administrative API endpoint to re-enable object summary retrieval. (#859)
    -   Restored the `/idp/openid/list-access-keys-bulk` and `/info-access-key` administrative API endpoints for managing OIDC access keys. (#857)
-   **Operations & Stability:**
    -   Reduced '503 Service Unavailable' errors for large multipart uploads in high-latency stretched cluster environments. (#861)
    -   Resolved a server crash that could occur when moving files to trash on a nearly full storage drive. (#896)
    -   Resolved a rare server crash that could occur during startup if metrics were collected before backend initialization was complete. (#917)
    -   Resolved an incompatibility with encrypted object metadata formats, ensuring reliable access to objects encrypted with different key versions. (#882, #885)
    -   Ensured successful deletion of non-versioned objects even when a storage node is offline and metadata quorum is temporarily lost. (#854)
    -   Improved the accuracy of drive health checks by independently monitoring read and write operations, ensuring faster detection of partially failing drives. (#848)
    -   Prevented a resource leak by ensuring background tasks (goroutines) are properly cleaned up after internal errors. (#892)
    -   Fixed a regression that caused batch catalog operations to fail due to premature timeouts. (#894)
    -   Enabled the successful re-import of the `site-replicator-0` service account to ensure site replication functions correctly after account removal. (#904)
-   **Configuration & Management:**
    -   Ensured that bucket storage quotas set with older software versions are correctly enforced after an upgrade. (#870)
    -   Refined license enforcement for stricter adherence to the `NotAfter` date. Trial licenses will now cease operation immediately upon expiration without a grace period. (#910)
    -   Reduced log noise by removing non-critical messages related to bucket quota parsing after server restarts. (#900)
    -   Reduced audit log volume by eliminating excessive entries generated by internal tiered object decommissioning operations. (#880)
    -   Improved error handling for Elasticsearch index creation to ensure failures are correctly detected and reported. (#888)
    -   Resolved a MinIO Client (mc) crash that occurred when rendering diagnostic data in a table format. (#864)
    -   Improved the diagnostic value of `mc admin trace` by displaying more accurate and descriptive handler names for API operations. (#912)