4.1.4 Release Note

[yiguolei]

4.1.3 release notes

New Features

AI & Search

• Support ANN indexes on Merge-on-Write tables (branch-4.1: [feature](fe) Support ANN indexes on MoW tables #67155)
• Support multimodal file embedding (branch-4.1: [Feature](multimodal_embed) Support multimodal(file) embed #66461)

Query & Execution

• Add adaptive global runtime filter tree publishing to support large cluster (branch-4.1: [feature](runtime-filter) Add adaptive global runtime filter tree publish #65599)
• Push down limits into CTE producers ([feature](fe) Push down limit into CTE producer  #63675)
• Add partition-filter SQL block rules ([feature](fe) Add partition filter sql block rule #62196)

Load & Streaming

• Add adaptive random-bucket load routing (branch-4.1: [feature](load) introduce adaptive random bucket load routing #65450)
• Support forwarding group-commit Stream Load requests ([fix](group commit) support forward group commit stream load #63594)
• Support compute-group routing and planning for Stream Load ([fix](load) support compute_group in FE stream load routing and planning #65571)
• Support OceanBase CDC streaming jobs ([feature](streaming-job) Support OceanBase CDC streaming jobs #65588)
• Detect PostgreSQL schema changes from Relation events ([improvement](streaming-job) Detect PostgreSQL schema changes from Relation events #64850)
• Support MySQL ADD and DROP schema changes in streaming jobs ([improve](streaming-job) Support MySQL ADD DROP schema changes #65325)

Cloud Native

• Add MetaService RPC rate limiting in FE ([feature](fe) Add meta service RPC rate limiting in FE #65694)
• Add memory governance for external metadata caches (branch-4.1: [feat](cache-memory) add external metadata cache memory governance #66717)
• Support SHOW COMPUTE GROUPS in non-cloud mode ([feature](compute group) Support show compute groups in non cloud mode #66697)
• Support tablet-level compaction through SQL ([opt](compaction) support tablet-level compaction through SQL #66611)
• Add resource group success-quorum checks ([feature](fe) Add resource group success quorum check #66751)

Lakehouse

• Support altering Iceberg and Paimon table properties (branch-4.1: [feature](catalog) Support altering Iceberg and Paimon table properties #66428)
• Support Alibaba Cloud OSS Tables REST catalog (branch-4.1: [feature](catalog) Support Alibaba Cloud OSS Tables REST catalog #66823)

Security & Authentication

• Add TLS support for internal Doris communication ([feature](tls) Add TLS framework #64016)
• Add internal HTTP authentication between FE and MetaService (branch-4.1: [improvement](fe) add fe_meta_auth_token for FE meta-service internal HTTP auth #66090)

Other

• Add queue_time_ms to audit logs (branch-4.1: [feat](audit log) add queue_time_ms in audit_log #66641)

Improvement

Query & Execution

• Optimize MD5 with an AVX2 batch path ([opt](function) speed up md5 with AVX2 batch path #63484)
• Improve percentile aggregate performance ([Opt](func) Improve AggFunc Percentile performance #62520)
• Optimize hash join probe-side output performance (branch-4.1: [improvement](be) Optimize hash join probe side output #65600)
• Optimize shuffle-key selection for Repeat decomposition ([improvement](repeat) Optimize shuffle key selection for Repeat decomposition #66532)
• Deduplicate grouping scalar functions in Repeat projections ([opt](repeat) deduplicate grouping scalar functions from projetion inrepeat #65880)
• Infer set-operation distinctness from NDV statistics ([improvement](fe) Infer set operator distinct by NDV #64618)
• Avoid underestimating NOT IN predicate row counts ([improvement](fe) Avoid under-estimating not-in predicate row count #66632)
• Add instance IDs to hash join profiles ([improvement](be) Add instance ID to hash join profiles #66097)
• Add spill-read deserialization timing ([fix](spill) Record the spill read deserialize timer #67041)
• Make random shuffle prefer local channels ([Exec](shuffle) random shuffle firstly use local channel to improve performance #59431)

Storage & Compaction

• Add remote-index byte metrics for the file cache ([improve](metrics) Export file cache remote index byte metrics #65398)
• Add queue-time metrics for delete-bitmap tasks (branch-4.1: [improvement](be) Add queue time for delete bitmap tasks #65523)
• Speed up file-cache LRU restoration at startup ([fix](file-cache): speed up LRU restore startup path #65174)
• Limit the file-cache LRU recorder shadow queue ([fix](filecache) Limit LRU recorder shadow queue size #64798)
• Skip redundant TTL scans for non-TTL tablets ([fix](filecache) skip redundant ttl scans for non-TTL tablets #65434)
• Add an option to disable file-cache writes during TopN lazy materialization (branch-4.1: [improvement](topn) Add option to skip file cache writes in lazy materialization #66414)
• Disable file-cache writes after a remote-scan threshold is reached (branch-4.1: [feature](file-cache) Disable cache writes after remote scan threshold #66479)
• Improve memory-limit refresh when cgroup settings change in serverless mode ([bugfix](memory) should update memory more quickly when user changes memory in cgroup directly in serveless mode  #66463)
• Backport nullable-column and projection optimizations ([opt](be) Backport nullable and projection optimizations to branch-4.1 #66586)

Load & Streaming

• Optimize nullable-string deserialization in CSV and text load ([opt](csv reader) optimize nullable string deserialization in CSV/text load hot path #64476)
• Backpressure asynchronous group commit by per-table WAL count ([fix](be) Backpressure async group commit by table WAL count #65362)

Cloud Native

• Add S3 rate-limit observability ([fix](s3) Add S3 rate limit observability #64038)
• Add warm-up job count metrics on BE (branch-4.1: [opt](cloud) Add warm up job count metric on BE #66136)
• Use rendezvous hashing for colocate-tablet placement ([improvement](cloud) Use rendezvous hashing for colocate tablet placement #64638)
• Add MetaService rate-limit dry-run observability ([improvement](cloud) Add dry-run observability for MetaService rate limiting #66969)
• Add cloud-tablet rebalancer metrics ([improvement](fe) Add cloud tablet rebalancer metrics #66576)
• Presize global cloud-tablet route sets ([improvement](fe) Presize global cloud tablet route sets #66447)
• Skip non-MoW tablets when calculating TopN delete-bitmap scores ([improve](cloud) Skip non-MOW tablets for get_topn_tablet_delete_bitmap_score #65964)

Lakehouse

• Cache credentials providers in the Hadoop S3A glue layer ([fix](glue) cache credentials provider in ConfigurationAWSCredentialsProvider #65165)
• Use credential-aware Hadoop FileSystem cache keys (branch-4.1: [fix](filesystem) Credential-aware hadoop FileSystem cache via doris.fs.cache.key #65586)
• Accelerate fixed-binary decimal decoding in Parquet ([opt](parquet) accelerate fixed binary decimal decoding #66379)
• Improve nullable selection planning for fragmented Parquet batches ([opt](parquet) Fuse fragmented nullable selection planning #66397)

Bugfix

AI & Search

• Fix crashes caused by null-query inverted-index predicates ([fix](be) Fix null query crash in inverted index predicates #65138)
• Split bound multi-segment inverted-index readers ([fix](inverted index) Split bound multi-segment readers #63138)
• Fix CLucene multi-segment readBlock handling ([fix](clucene) Fix clucene multi-segment readBlock #66736)
• Reject inverted-index terms containing embedded NUL bytes ([fix](inverted index) update clucene to reject terms containing an embedded NUL byte #67179)
• Disable the DSL cache for score queries ([fix](be) Disable DSL cache for score queries #65436)
• Validate nested Variant MATCH predicates ([fix](reverted index) Validate nested variant MATCH predicates #66207)
• Fix index lookup after indexes are removed ([fix](be) Key the index lookup map by each surviving index in remove_index #66316)
• Preserve atomicity when appending nested Variant values ([fix](variant) Preserve nested Variant append atomicity #66421)
• Recognize nullable array elements as NestedGroup types ([fix](variant) Recognize nullable array elements as NestedGroup types #66196)
• Prefer sparse Variant fields over root values ([fix](variant) Prefer sparse variant fields over root value #65660)
• Serve shredded Variant leaves without rebuilding the root (branch-4.1:[fix](variant) Serve shredded Variant leaves without rebuilding the root #66941)
• Preserve NestedGroup access paths through explode (branch-4.1: [fix](variant) Preserve NestedGroup access paths through explode #67022)
• Avoid recursion when the ANN IVF-list cache is absent ([fix](be) Read instead of recursing when the ANN IVF list cache is absent #67024)

Query & Execution

• Fix analytic partition keys overflowing before conversion to 64-bit columns ([bug](analytic) fix partition by column when it overflow haven't convert to column64 #65240)
• Guard execution-profile access during concurrent teardown ([branch-4.1][fix](profile) guard execution profile access #65442)
• Fix duplicate join nodes produced by PushDownAggThroughJoinOnPkFk ([fix](fe) Fix duplicate join nodes in PushDownAggThroughJoinOnPkFk tree reconstruction #65172)
• Guard UniqueFunction handling in filter and TopN pushdown rules ([fix](nereids) Guard UniqueFunction in multiple filter/topn pushdown rules #62742)
• Enable pre-aggregation for multi-argument aggregates with all-key distinct inputs ([fix](preagg) Enable pre-aggregation for multi-argument aggregate functions with all-key distinct inputs #65846)
• Fix missing decimal precision and scale for legacy decimal data ([fix](be) Fix missing decimal precision and scale for legacy decimal data #65419)
• Materialize constant columns before block merge ([fix](be) Materialize const columns before block merge #65770)
• Use serialized hash keys for complex types ([fix](be) Use serialized hash keys for complex types #66777)
• Fix blocking-queue waiter accounting ([fix](be) Fix waiter accounting in blocking queue #65827)
• Fix unsafe eager-aggregation nullability conversions (branch-4.1: [fix](eager-agg) null-to-non-null safety check and remove redundant normalize calls in eager aggregation #66208)
• Fix duplicate aggregate functions pushed through projects ([fix](eager-agg) Handle duplicate aggregate functions pushed through projects #66531)
• Preserve NULL in pushed-down CHAR MIN and MAX ([fix](exec) preserve NULL in pushed-down CHAR MIN/MAX #65952)
• Fix invalid semi-join transpose when the lower join is a mark join ([fix](mark join) Fix invalid semi-join transpose when the bottom join is a mark join #66574)
• Drop invalid functional dependencies from the outer side of joins ([fix](fd)drop Function dependencies from join outer side #65982)
• Prevent unsafe CTE runtime-filter pushdown ([fix](fe) Prevent unsafe CTE runtime filter pushdown #65247)
• Suppress invalid unique constraints in scan functional-dependency derivation ([fix](fd) Suppress invalid unique constraints in scan FD derivation #66801)
• Fix TIMESTAMPTZ nullable type equality ([fix](be) Fix TIMESTAMPTZ nullable type equality #66722)
• Fix TIMESTAMPTZ values in COALESCE ([fix](be) Fix TIMESTAMPTZ values in COALESCE #66689)
• Preserve TIMESTAMPTZ user-variable types ([fix](user variable) Preserve TIMESTAMPTZ user variable type #66833)
• Handle DST fallback in timestamp-cast monotonicity ([fix](cast) Handle DST fallback in timestamp cast monotonicity #65903)
• Accept spaced TIMESTAMPTZ offsets in range-partition bounds (branch-4.1: [fix](fe) Accept spaced TIMESTAMPTZ offsets in range partition bounds  #66292)
• Return Arrow and Thrift DATETIME values as naive timestamps ([fix](exec) Return DATETIME as naive Arrow timestamp  #67027, [fix](exec) Return Thrift DATETIME as naive Arrow timestamp #67232)
• Correct the Flight SQL GetTables schema and TIMESTAMPTZ Arrow reader ([fix](arrow) correct the Flight SQL GetTables schema and the TIMESTAMPTZ arrow reader #66344)
• Fix count_substrings tail overmatching ([fix](function) prevent count_substrings tail overmatch #63215)
• Avoid ICU default-locale contention in functions ([fix](function) avoid ICU default-locale contention #66440)
• Keep multi-match regex cache ownership alive ([Fix](func) Keep multi-match regex cache owner alive #66909)
• Prevent explode_bitmap crashes when output size exceeds INT_MAX ([fix](be) fix BE core due to explode_bitmap size exceeding INT_MAX #66034)
• Skip CHAR payload checks for NULL rows ([fix](be) Skip CHAR payload checks for NULL rows #67043)
• Restore null-literal return-type handling ([fix](be) Restore null literal return type handling #66280)
• Stop strict-mode string casts from attaching NULL rows to the next value ([fix](serde) Stop strict-mode string casts gluing NULL rows onto the next value #66952)
• Materialize constant virtual columns before caching ([fix](virtual slot) Materialize constant virtual columns before caching #67112)
• Handle case-insensitive auto-partition function arguments ([fix](function) Handle case-insensitive auto partition arguments #67121)
• Preserve nested paths during lazy row-ID fetch ([fix](be) Preserve nested paths for lazy rowid fetch #67205)
• Handle empty row-ID fetch RPC failures in TopN ([fix](topn) Handle empty rowid fetch RPC failures #66443)
• Key UDF class caches by function ID and clean them up in cloud mode ([Fix](udf) Key UDF class cache by function ID and enable cleanup in cloud mode #67046)

Storage & Compaction

• Preserve shortest round-trip representations for FLOAT and DOUBLE ZoneMaps ([fix](storage) fix DOUBLE/FLOAT zone-map  #65302)
• Fix floating-point equality during Parquet pruning ([fix](be) Preserve floating-point equality in Parquet pruning #66470)
• Fix sparse vertical-compaction destination offsets ([fix](compaction) Fix sparse vertical compaction destination offset #66306)
• Reject transaction preparation on tablets that are shutting down ([fix](storage) Reject prepare txn on shutdown tablet #66448)
• Clean up spill directories during query teardown (branch-4.1: [fix](be) Clean up spill directories during query teardown #66458)
• Avoid scan-executor use-after-free during shutdown ([fix](be) Avoid scan executor shutdown use after free #65220)
• Pass tablet IDs explicitly through file-reader and packed-file cache contexts ([fix](filecache) pass tablet_id through FileReaderOptions instead of parsing from path #61683, branch-4.1: [fix](packed-file) set tablet_id in CacheContext when writing small file to file cache #65701)
• Avoid converting segment-cache blocks to index-cache blocks ([fix](be) remove changing segment cache blocks to index type #65905)
• Fix BE memory leaks after failed schema-change jobs ([fix](bvar) Fix bvar leak after schema change job in be failed #56207)
• Fix PublishVersionDaemon thread-pool leaks ([fix](memory) fix ThreadPoolExecutor leak in PublishVersionDaemon  #60720)
• Make nullable hash-state updates exception-safe ([fix](be) Backport nullable hash state exception safety to branch-4.1 #66517)
• Synchronize access to load error logs ([fix](be) Synchronize load error log access #66250)
• Fix cache lookup failures after an OLAP partition is dropped ([fix](cache) fix NPE when a partition of olap table has been dropped #65889)
• Skip file cache for HTTP chunk-response readers ([fix](be) Skip file cache for http chunk-response readers #66932)

Load & Streaming

• Respect profile levels for load-writer details (branch-4.1: [fix](load) respect profile level for writer details #64978)
• Fix Arrow Stream Load with uppercase column names (branch-4.1: [fix](load) Fix Arrow stream load with uppercase columns #65617)
• Fail fast when a backend restarts during INSERT ([fix](transaction) Fail fast when a backend restarts during insert #65525)
• Skip decommissioning backends during load routing (branch-4.1: [fix](fe) Skip decommissioning BE for load #65406)
• Correct quorum participants for incremental load streams ([fix](load) correct quorum participants for incremental streams #66016)
• Fix stream-receiver leaks racing with close_load ([fix](load_stream) fix receiver-side stream leak race on close_load #65584)
• Fix broker-load label self-conflicts during pending retries ([fix](load)(regression-test) fix broker load label self-conflict on pending retry; deflake P2 suites #66469)
• Recover Broker Load jobs whose transactions are already visible ([fix](job) Recover broker load from visible transaction #66987)
• Clear transaction IDs after aborting Broker Load transactions ([fix](job) Clear transaction ID after aborting broker load transaction #66884)
• Encode CDC Stream Load records as UTF-8 ([fix](streaming-job) Encode CDC stream load records as UTF-8 #66771)
• Prevent MySQL CDC data loss on non-GTID keepalive reconnects ([fix](streamingjob) Prevent MySQL CDC data loss on keepalive reconnect for non-GTID #66998)
• Isolate schemas across CDC snapshot splits ([fix](streaming-job) Isolate schemas for CDC snapshot splits #65645)
• Keep CDC end offsets consistent with current progress ([fix](fe) Keep CDC end offset consistent with current progress #65688)
• Normalize MySQL JDBC URLs used by streaming jobs ([fix](fe) Normalize MySQL JDBC URLs for streaming jobs #65647)
• Avoid applying a global BE authentication precheck to Stream Load ([fix](load) Avoid global BE auth precheck for stream load #66629)
• Report Stream Load size limits in MiB ([fix](be) Report stream load size limits in MiB #66224)
• Remove the unsupported Workload Policy action that mutates session variables ([fix](workload policy)Remove set session variable workload policy action #64856)

Cloud Native

• Retry MetaService too busy responses ([fix](fe) Retry meta service too busy responses #65378)
• Set a recycler S3 request timeout to avoid slow DeleteObjects failures ([fix](cloud) set recycler S3 client requestTimeoutMs to avoid curl-28 on slow DeleteObjects (mirror #49315) #64758)
• Preserve resource IDs when recycling empty rowsets ([fix](cloud) Preserve resource IDs when recycling empty rowsets #65862)
• Skip packed-slice deletion while recycling rowsets (branch-4.1: [fix](recycler) skip packed slice deletion during recycle rowsets #65533)
• Fix local and remote tablet-size semantics in schema views ([fix] (cloud) Fix local/remote tablet size semantics in schema views #60887)
• Normalize storage and replica fields in SHOW PARTITIONS ([fix](cloud) normalize SHOW PARTITIONS display for storage and replica #60871)
• Align colocate proc output and tablet-health reporting in cloud mode ([fix](cloud) Align colocate proc output and tablet health in cloud mode #60944)
• Exclude decommissioning backends from cloud backend selection ([fix](cloud) Exclude decommissioning BE from BE selection #65221)
• Prevent urgent loads from preempting schema-change locks ([fix](cloud) Prevent urgent load from preempting schema change lock #66082)
• Skip unnecessary MetaService tablet updates after table-property changes ([fix](cloud) Skip unnecessary update tablet to ms when alter table property #65983)
• Prevent older clients from treating unknown MetaService status codes as success ([fix](cloud) Prevent old clients from decoding unknown MetaServiceCode as OK #64148)
• Fail closed on unknown MetaService response codes ([fix](cloud) Fail closed on unknown MetaService response codes #66364)
• Skip rowset promotion for incomplete lazy commits (branch-4.1: [fix](cloud) Skip BE rowset promotion for incomplete lazy commits #66253)
• Track deleted-instance recycling status and retain instance tombstones ([fix](cloud) Track deleted instance recycle status #66519, [fix](cloud) Retain instance tombstones after recycle cleanup #66870)
• Recycle delete-bitmap packed files with correct blob reads and reference counts ([fix](cloud) recycler decrement_delete_bitmap_packed_file_ref_counts method should read delete bitmaps as blobs #66728, [fix](cloud) Decrement delete bitmap packed file ref counts when recycling tablets #66919)
• Only mark prepared rowsets before recycling ([fix](cloud) Only mark prepare rowsets before recycle #65550)
• Skip versioned delete-bitmap cleanup for non-MoW tablets ([fix](cloud) Skip versioned delete bitmap cleanup for non-MoW tablets #67084)
• Separate HTTP encoding for recycler job and check keys ([fix](cloud) Separate HTTP encoding for job recycle and check keys #67243)
• Drop stale CloudReplica routes after compute groups are deleted ([fix](cloud) Drop CloudReplica route entries of dropped compute groups #66984)
• Fix SchemaVariablesScanner crashes on incomplete FE responses ([fix](cloud) Fix SchemaVariablesScanner crash due to incomplete FE response #65994)
• Use a bthread-aware shared mutex for tablet-header locks (branch-4.1: [fix](cloud) Use bthread-aware shared mutex for tablet header lock #66020)
• Fix schema-change alter-version capture in cloud mode ([fix](schema-change) Use actual schema change alter version in cloud capture #62506)
• Restrict runtime FE configuration updates to root in cloud mode ([fix](config) Restrict frontend config updates to root in cloud mode #66478)
• Resolve conflicting default replica properties ([fix](replica) Resolve conflicting default replica properties #65836)
• Exclude zero-size tablets from backend balancing ([fix](balance) do not pick tablets whose data size is zero for BE balance #66499)
• Inspect and abort failed conflicting transactions during cloud upgrades ([fix](cloud) CloudUpgradeMgr inspect and abort failed conflict txns while waiting #60830)
• Cancel rebuilt virtual-compute-group warm-up jobs when the group is dropped ([fix](fe) Cancel rebuilt VCG warm up jobs on drop #65426)
• Fix rollback handling when adding partitions ([fix](partition) handle rollback failure when adding partition #64650)

Lakehouse

• Fix MaxCompute IN predicate pushdown polarity ([fix](maxcompute)Fix MaxCompute IN predicate pushdown polarity. #65083)
• Improve MaxCompute catalog validation ([fix](fe) Improve MaxCompute catalog validation #64119)
• Preserve external partition metadata in multi-catalog queries (branch-4.1: [fix](multi-catalog) Preserve external partition metadata #66012)
• Invalidate stale Hive file-cache entries after partition refresh ([fix](hive) invalidate stale file cache on partition refresh when partition cache misses #65334)
• Safely publish Hadoop catalog properties ([fix](catalog) safely publish Hadoop properties #66392)
• Resolve missing remote JDBC table names during schema lookup ([fix](jdbc) Resolve missing remote table names in schema lookup #65718)
• Harden JDBC driver URL validation and remove the file-upload HTTP API (branch-4.1: [fix](jdbc) Harden JDBC driver URL validation and remove file-upload HTTP API #67149)
• Decode ORC timestamps through the standard SerDe path ([fix](orc) decode timestamp through serde #64807)
• Preserve nested nullability in Parquet predicate scans ([fix](parquet) Handle nested nullability in predicate scans #66799)
• Safely prune nested Parquet columns with Bloom filters ([fix](be) Safely prune nested Parquet columns with Bloom filters #66471)
• Fix Iceberg Merge-on-Read crashes with row-ID lazy reads ([Bug](iceberg) fix BE core when reading MOR table with rowid + lazy read #66506)
• Avoid BE aborts when evaluating Iceberg partition predicates (branch-4.1: [fix](iceberg) Avoid BE abort on partition predicates #66835)
• Use the split file format when scanning Iceberg tables ([bug](iceberg) use split file format for iceberg scan #65760)
• Delete external data files after failed writes to prevent orphan files ([fix](external-write) Delete the actual data file on write failure to avoid orphan files #64678)
• Preserve Paimon partition metadata across splits (branch-4.1: [fix](paimon) preserve partition metadata across splits #65581)
• Handle special characters in Paimon partition values ([fix](paimon) Handle special characters in partition values #65904)

Security & Authentication

• Require authorization and an explicit configuration gate for _stream_load_forward ([branch-4.1][fix](security) Add auth and config gate for _stream_load_forward endpoint #65377)
• Mask Kafka Routine Load sensitive properties ([fix](fe) Mask Kafka routine load sensitive properties #64786)
• Hide Stream Load tokens and authentication data from BE information endpoints ([enhance](auth)Hidden the token info during StreamLoad in BE info #60656, [enhance](auth)Hidden the auth info during StreamLoad in BE info #59743)
• Escape audit-log record separators to prevent row forgery ([fix](audit) escape 0x1F/0x1E in audit_log stream load to prevent row forgery #66580)
• Mask sensitive fields in encryption-key schema tables ([fix](security) Mask sensitive fields in encryption keys schema table #66834)
• Mask private-key passwords in backend configuration ([fix](security) Mask private key password in backend configuration #66836)
• Mask credentials in authentication and Stream Load logs (branch-4.1: [fix](auth) Mask credentials in authentication and Stream Load logs #66917)
• Stop writing Arrow Flight bearer tokens to fe.log (branch-4.1: [fix](arrow-flight) Stop writing bearer tokens to fe.log #67146)

Materialized Views

• Fix scan-materialization pattern-context checks ([fix](mv) Fix scan materialization pattern context check #65414)
• Distinguish partition compensation from UNION ALL rewrite ([fix](mv) Distinguish partition compensation from UNION ALL rewrite #66445)
• Reduce master FE CPU usage when serializing MTMV task TVF information ([fix](mtmv) reuse Gson instance in MTMVTask.getTvfInfo to avoid Master FE CPU spike #66792)
• Optimize MTMV partition-lineage checks ([improvement](mtmv) Optimize MTMV partition lineage check #63899)
• Avoid invalid slot casts in MV null-reject compensation ([fix](mtmv) Avoid invalid slot cast in MV null-reject compensation #66613)
• Refresh MTMVs after excluded trigger tables change ([fix](mtmv) Refresh MTMV after excluded trigger tables change #64041)

Other

• Fix Arrow Flight SQL direct-memory leaks in prepared statements ([fix](arrow-flight-sql) Close temporary VectorSchemaRoot in createPreparedStatement to fix FE direct memory leak #65311)
• Fix task finish timestamps when scheduled jobs fail ([fix](job scheduler) Record task finish time on execution failure #66232)
• Guard query-instance metrics before metric repository initialization ([fix](metrics) Guard query instance metric before metric repo init #62762)
• Remove query profiles from HDFS file readers (branch-4.1: [fix](hdfs) Remove query profile from HDFS file reader #67293)