# TiKV v8.5.4

2025-11-27 · [Original release notes](https://github.com/tikv/tikv/releases/tag/v8.5.4) · [Full summary](https://ratatosk.io/en/releases/tikv/v8.5.4)

Breaking changes 5 · Deprecations 1 · Important fixes 42 · Enhancements 11 · Dependency updates 1 · Other changes 50

## Check before upgrading

### Prerequisites

- **Smooth upgrades from newly created v8.5.3 clusters**: This release changes defaults and behavior for system variables and configuration parameters.
  > Newly created v8.5.3 clusters can be smoothly upgraded to v8.5.4. However, v8.5.4 introduces several default value changes and behavior adjustments for system variables and configuration parameters. Before upgrading, pay attention to the following:

  https://docs.pingcap.com/tidb/stable/release-8.5.4
- **Upgrade guidance for customized TiFlash and TiKV compaction configurations**: Most changes are safe for routine upgrades; clusters with these custom configurations need careful review.
  > - Most changes are safe for routine upgrades. However, if your cluster has undergone performance tuning, such as customized TiFlash and TiKV compaction configurations, read this section carefully.

  https://docs.pingcap.com/tidb/stable/release-8.5.4

## Breaking changes (5)

### `tidb_replica_read` limited to read-only SQL

- Affects: SQL statements using `tidb_replica_read` (Inferred)

> tidb_replica_read Modified Starting from v8.5.4, this variable only takes effect on read-only SQL statements. This improves data read safety and reduces overlaps with other features. #62856 @you06

https://docs.pingcap.com/tidb/stable/release-8.5.4

### CPU-based default for `server.grpc-raft-conn-num`

- Affects: TiKV setups using the default for `server.grpc-raft-conn-num`

Previously, the default was 1; it now follows `MAX(1, MIN(4, CPU cores / 8))`, adapting the connection count to CPU core count.

> TiKV server.grpc-raft-conn-num Modified Changes the default value from 1 to MAX(1, MIN(4, CPU cores / 8)) , which enables gRPC-related thread settings to adapt to the number of CPU cores. When the number of CPU cores equals or exceeds 32, the maximum default number of connections is 4. #18806 @LykxSassinator

https://docs.pingcap.com/tidb/stable/release-8.5.4

### TiKV changes the `server.grpc-concurrency` default

- Affects: TiKV setups using the default for `server.grpc-concurrency`

The default was 5; setups still using it now get `grpc-raft-conn-num * 3 + 2` instead, adapting gRPC-related thread settings to the number of CPU cores.

> TiKV server.grpc-concurrency Modified Changes the default value from 5 to grpc-raft-conn-num * 3 + 2 , which enables gRPC-related thread settings to adapt to the number of CPU cores. #18806 @LykxSassinator

https://docs.pingcap.com/tidb/stable/release-8.5.4

### `max_execution_time` enforcement fixed for `SELECT FOR UPDATE` statements

- Affects: setups with `max_execution_time` configured for `SELECT FOR UPDATE` statements

Before this fix, `max_execution_time` did not take effect when a `SELECT FOR UPDATE` statement ran.

> - Fix the issue that max_execution_time does not take effect for SELECT FOR UPDATE statements #62960 @ ekexium

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Bucket splitting stops when bucket reporting is disabled

- Affects: setups with bucket reporting disabled and bucket splitting in use (Inferred)

Before this fix, bucket splitting still worked when bucket reporting was disabled.

> - Fix the issue that splitting buckets still works even after bucket reporting is disabled #9726 @ bufferflies

https://docs.pingcap.com/tidb/stable/release-8.5.4

## Deprecations (1)

### TiKV automatic-compaction settings replaced by `gc.auto-compaction`

- Affects: TiKV configurations using the `region-compact-*` items or `gc.auto-compaction` settings (Inferred)
- Removed in: not stated

The `region-compact-*` items were the existing settings for this behavior; the new group controls automatic compaction.

> TiKV - region-compact-check-interval
> - region-compact-check-step
> - region-compact-min-tombstones
> - region-compact-tombstones-percent
> - region-compact-min-redundant-rows
> - region-compact-redundant-rows-percent
> Deprecated These configuration items are replaced by the gc.auto-compaction configuration group, which controls the automatic compaction behavior. #18727 @v01dstar
> TiKV gc.auto-compaction configuration group: - gc.auto-compaction.check-interval
> - gc.auto-compaction.tombstone-num-threshold
> - gc.auto-compaction.tombstone-percent-threshold
> - gc.auto-compaction.redundant-rows-threshold
> - gc.auto-compaction.redundant-rows-percent-threshold
> - gc.auto-compaction.bottommost-level-force
> Newly added This configuration group controls the automatic compaction behavior. #18727 @v01dstar

https://docs.pingcap.com/tidb/stable/release-8.5.4

## Important fixes (42)

### Row count overestimation corrected for long spans

- Affects: row count estimates spanning months or years

Those estimates could be significantly too high.

> - Fix the issue that row count estimates across months or years can be significantly overestimated #50080 @ terry1purcell

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Incorrect regular-expression query results fixed for `INFORMATION_SCHEMA`

- Affects: `INFORMATION_SCHEMA` queries

Queries using regular expressions against `INFORMATION_SCHEMA` tables might return incorrect results.

> - Fix the issue that queries using regular expressions against INFORMATION_SCHEMA tables might return incorrect results #62347 @ River2000i

https://docs.pingcap.com/tidb/stable/release-8.5.4

### `ADMIN ALTER DDL JOBS` parameter display fixed

- Affects: setups that dynamically modify parameters and use `ADMIN ALTER DDL JOBS` statements

After parameters were dynamically modified, `ADMIN ALTER DDL JOBS` statements displayed incorrect parameter values.

> - Fix the issue that ADMIN ALTER DDL JOBS statements display incorrect parameter values after dynamically modifying parameters #63201 @ fzzf678

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Incorrect `ATAN2` results fixed

- Affects: queries using the `ATAN2` function

Using `ATAN2` could produce incorrect results; this release fixes that behavior.

> - Fix the issue that using the ATAN2 function might produce incorrect results #60093 @ guo-shaoge

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Variable restoration after `set_var` hints with bindings fixed

- Affects: bindings using the `set_var` hint

When the `set_var` hint was used with bindings, variables could remain unrestored and keep settings other than their original values. They are now restored to their original settings in this case.

> - Fix the issue that using the set_var hint with bindings prevents variables from being restored to their original settings #59822 @ wddevries

https://docs.pingcap.com/tidb/stable/release-8.5.4

### [Silent failure] Incorrect results from outer-to-inner join rewrites fixed

- Affects: queries using outer-to-inner join rewrites

When an outer join was rewritten to an inner join, the rewrite could produce incorrect results. This release fixes those results.

> - Fix the issue that rewriting an outer join to an inner join might produce incorrect results #61327 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-8.5.4

### [Silent failure] Incorrect global-index results fixed

- Affects: setups using global indexes

In some cases, the global index could return incorrect results. Those incorrect results are fixed in this release.

> - Fix the issue that the global index might return incorrect results in some cases #61083 @ Defined2014

https://docs.pingcap.com/tidb/stable/release-8.5.4

### [Silent failure] `character_set_results` character replacement fixed

- Affects: setups using `character_set_results`

Previously, affected characters were truncated instead of replaced. They are now replaced rather than truncated.

> - Fix the issue that character_set_results truncates incorrect characters instead of replacing them #61085 @ xhebox

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Incorrect TSO HTTP API results fixed

- Affects: setups using the TSO HTTP API `/config` or `/members` endpoints

Before this release, these endpoints returned incorrect results.

> - Fix the issue that the TSO HTTP API /config and /members return incorrect results #9797 @ lhy1024

https://docs.pingcap.com/tidb/stable/release-8.5.4

### TiFlash RU-consumption statistics inflation fixed

- Affects: TiFlash setups that generate statistics for RU consumption (Inferred)

When TiFlash generated statistics for RU consumption, the reported values were inflated.

> - Fix the issue that TiFlash generates inflated statistics for RU consumption #10380 @ JinheLin

https://docs.pingcap.com/tidb/stable/release-8.5.4

### [Silent failure] `FLOOR()` and `CEIL()` `DECIMAL` results fixed

- Affects: TiFlash setups using `FLOOR()` or `CEIL()` with `DECIMAL` parameters

When called with a `DECIMAL` parameter, these functions could return incorrect results; they now return corrected results.

> - Fix the issue that the FLOOR() and CEIL() functions might return incorrect results when their parameters are of the DECIMAL type #10365 @ ChangRui-Ryan

https://docs.pingcap.com/tidb/stable/release-8.5.4

### `mysql.stats_meta` count restoration fixed

- Affects: backups made with `--checksum` set to false

After restore, the `count` column in `mysql.stats_meta` could become 0. This issue is fixed in this release.

> - Fix the issue that setting --checksum to false during backup might cause the count column in the mysql.stats_meta table to become 0 after restore #60978 @ Leavrth

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Log backup zstd compression restored

- Affects: TiKV setups using log backup with the default zstd fastest setting

When log backup used the default algorithm expected to be zstd fastest, `async-compression` v0.4.12's negative default level disabled or made zstd compression ineffective.

> - tikv/tikv#18836
> - Fix the issue that zstd compression in log backup does not take effect, causing the output to remain uncompressed #18836 @ 3pointer

### Azure Blob backup flush slowdown fixed

- Affects: setups backing up to Azure Blob storage with uploads reading from a non-memory source

During backups to Azure Blob Storage, a 400 MiB flush could take about 17 minutes and use a full CPU core.

> - tikv/tikv#18410
> - Fix the issue that the flush operation occasionally becomes slow when backing up data to Azure Blob Storage #18410 @ YuJuncen

### Compaction scoring uses GC safe points

- Affects: TiKV setups that calculate compaction scores in relation to the GC safe point (Inferred)

MVCC redundancy is calculated using GC safe points and RocksDB statistics, improving the efficiency and accuracy of compaction scoring.

> - tikv/tikv#18697
> - Calculate MVCC redundancy based on GC safe points and RocksDB statistics to improve the efficiency and accuracy of compaction #18697 @ v01dstar

### Log backup region observation loss fixed

- Affects: log backup setups

When a region became leader while a new log backup task was initializing, reordered `Start` and `Stop` observe messages could leave it unobserved, causing progress to stop advancing with a stale observe ID error. The observer now retains observation of the region in this situation.

> - tikv/tikv#18243
> - Fix the issue that log backup observer might lose the observation on a Region, causing the log backup progress to fail to advance #18243 @ Leavrth

### TiKV CDC assertion panic fixed

- Affects: setups using TiKV CDC

On TiKV v8.5.1, a CDC processing path could hit an assertion with mismatched timestamps, which could cause TiCDC to panic.

> - tikv/tikv#18498
> - Fix the issue that assertion errors in TiKV might cause TiCDC to panic #18498 @ tharanga

### CDC region-subscription panic fixed

- Affects: setups using CDC

CDC could theoretically panic if a region was subscribed between a key rollback and the prewrite entry; the issue report said this had not occurred in reality.

> - tikv/tikv#19048
> - Fix the issue that TiCDC might panic when encountering rollback and prewrite entries during Region subscription #19048 @ 3AceShowHand

### ANALYZE support for index-related DDL

- Affects: TiDB setups with `tidb_stats_update_during_ddl` enabled

TiDB runs ANALYZE before a new or reorganized index becomes visible, helping avoid inaccurate optimizer estimates and plan changes while statistics are unavailable. The variable is OFF by default, and enabling it can increase ADD INDEX execution time on large tables.

> - Support ANALYZE embedded in DDL statements #57948 @ terry1purcell @ AilinKid
> This feature applies to the following types of DDL statements:
> - DDL statements that create new indexes: ADD INDEX
> - DDL statements that reorganize existing indexes: MODIFY COLUMN and CHANGE COLUMN
> When this feature is enabled, TiDB automatically runs an ANALYZE (statistics collection) operation before the new or reorganized index becomes visible to users. This prevents inaccurate optimizer estimates and potential plan changes caused by temporarily unavailable statistics after index creation or reorganization.
> For more information, see documentation .

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Automatically recoverable BR backup errors logged at WARN

- Affects: BR backup operations

During conditions such as `backup snapshot failed` and `backup writer failed`, these automatically recoverable errors were previously logged at `ERROR` despite not being error-level conditions and requiring manual operation.

> - Change the log level of certain automatically recoverable errors in the BR module from ERROR to WARN to reduce unnecessary alerts #18493 @ YuJuncen
> - tikv/tikv#18493

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Certain TiKV errors logged at `WARN`

- Affects: TiKV setups that emit the affected errors

When these errors occur, they were previously logged at `ERROR`; logging them at `WARN` reduces unnecessary alerts.

> - Change the log level of certain TiKV errors from ERROR to WARN to reduce unnecessary alerts #18745 @ exit-code-1

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Incorrect short-path processing in `TRUNCATE()` fixed

- Affects: queries that use the short path in `TRUNCATE()` (Inferred)

> - Fix the issue that the short path in the TRUNCATE() function is incorrectly processed #57608 @ xzhangxian1008

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Spilled-file cleanup fixed for local temporary space quota errors

- Affects: setups that spill files

When the Out Of Quota For Local Temporary Space error was triggered, spilled files might not be fully deleted.

> - Fix the issue that spilled files might not be fully deleted when the Out Of Quota For Local Temporary Space error is triggered #63216 @ xzhangxian1008

https://docs.pingcap.com/tidb/stable/release-8.5.4

### TiDB error reporting for failed PD timestamp retrieval fixed

- Affects: TiDB setups that get timestamps from PD

Previously, when TiDB failed to get a timestamp from PD, it did not return an error.

> - Fix the issue that TiDB does not return an error when it fails to get a timestamp from PD #58871 @ joechenrh

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Owner and non-owner TiDB query result mismatch fixed for `MODIFY COLUMN`

- Affects: `MODIFY COLUMN` statements across owner and non-owner TiDB instances

During `MODIFY COLUMN` statements, query results differed between owner and non-owner TiDB instances.

> - Fix the issue that query results differ between the owner TiDB and non-owner TiDB instances during executing MODIFY COLUMN statements #60264 @ tangenta

https://docs.pingcap.com/tidb/stable/release-8.5.4

### GC safe point advancement fixed for index creation in transactions

- Affects: transactions that add an index

The GC safe point did not advance when an index was added within a transaction.

> - Fix the issue that the GC safe point does not advance when adding an index within a transaction #62424 @ wjhuang2016

https://docs.pingcap.com/tidb/stable/release-8.5.4

### DXF task cancellation exit failure fixed

- Affects: setups using DXF

When a DXF task was canceled while other tasks were running, those tasks failed to exit.

> - Fix the issue that after the DXF task is canceled, other tasks fail to exit #63927 @ D3Hunter

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Apply operator plan generation failure fixed

- Affects: setups with Apply operator concurrency enabled using `tidb_enable_parallel_apply = on`

With `tidb_enable_parallel_apply = on`, plan generation failed because the Apply operator concurrency path was missing a clone implementation.

> - Fix the issue that enabling the Apply operator concurrency ( tidb_enable_parallel_apply = on ) causes plan generation failure due to a missing clone implementation #59863 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Case-insensitive behavior of the `ONLY_FULL_GROUP_BY` check fixed

- Affects: The `ONLY_FULL_GROUP_BY` check (Inferred)

When the check handled inputs whose behavior depended on case, it treated them case-insensitively. That behavior is fixed, but the note does not specify the replacement behavior.

> - Fix the issue that the ONLY_FULL_GROUP_BY check is case-insensitive #62672 @ AilinKid

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Internal panic on certain queries fixed

- Affects: setups that execute the affected queries

Executing certain queries could trigger an internal panic; the note does not identify which queries.

> - Fix the issue that executing certain queries might cause an internal panic #58600 @ Defined2014

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Incorrect global-index reads during `ALTER PARTITION` fixed

- Affects: setups using global indexes and performing `ALTER PARTITION` operations

During certain `ALTER PARTITION` operations, the global index could return incorrect data. Those reads are fixed in this release.

> - Fix the issue that the global index might read incorrect data during certain ALTER PARTITION operations #64084 @ mjonss

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Resource Manager token allocation fix

- Affects: setups using the Resource Manager (Inferred)

When the Resource Manager incorrectly allocated tokens, queries could get stuck.

> - Fix the issue that the Resource Manager incorrectly allocates tokens, causing queries to be stuck #9455 @ JmPotato

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Placement Rules after PD leader switches fixed

- Affects: setups using Placement Rules

When the PD leader switched, Placement Rules did not take effect.

> - Fix the issue that Placement Rules do not take effect after PD leader switches #9602 @ okJiang

https://docs.pingcap.com/tidb/stable/release-8.5.4

### PD parsing of large scientific-notation values fixed

- Affects: TTL-related configurations containing large numeric values in scientific notation

When PD encountered large numeric values in scientific notation, it could fail to parse them, so some TTL-related configurations did not take effect.

> - Fix the issue that PD might fail to parse large numeric values in scientific notation, which consequently causes some TTL-related configurations to not take effect #9343 @ lhy1024

https://docs.pingcap.com/tidb/stable/release-8.5.4

### TiFlash out-of-memory condition fixed

- Affects: TiFlash using disaggregated storage and compute

When slow queries existed, TiFlash could run out of memory. This condition is fixed in this release.

> - Fix the issue that TiFlash might encounter OOM when slow queries exist under the disaggregated storage and compute architecture #10278 @ JaySon-Huang

https://docs.pingcap.com/tidb/stable/release-8.5.4

### TiFlash indefinite retries with S3 fixed

- Affects: TiFlash deployments using disaggregated storage and compute with S3

During a network partition between TiFlash and S3, TiFlash could retry indefinitely. That indefinite-retry behavior is fixed.

> - Fix the issue that TiFlash might retry indefinitely when a network partition occurs between TiFlash and S3 under the disaggregated storage and compute architecture #10424 @ JaySon-Huang

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Log truncation on file-deletion failure fixed

- Affects: log operations involving file deletion (Inferred)

When file deletion failed, log truncation might occur.

> - Fix the issue that log truncate might occur when file deletion fails #63358 @ YuJuncen

https://docs.pingcap.com/tidb/stable/release-8.5.4

### BR restore failure risk reduced for bandwidth-limited S3-compatible storage

- Affects: S3-compatible storage services with bandwidth limits enabled

Stalled-stream protection could kill a download request, which could then restart from the beginning and fail with a “streaming error.” This release reduces the chance of that restore failure but does not eliminate it.

> - Reduce the possibility of BR failing to restore data from S3-compatible storage services when the bandwidth limits on these services are enabled #18846 @ kennytm
> - tikv/tikv#18846

https://docs.pingcap.com/tidb/stable/release-8.5.4

### TiCDC column-type partition dispatcher panic fixed

- Affects: TiCDC column-type partition dispatchers with virtual columns

> - Fix a panic that could occur when configuring a column-type partition dispatcher that includes virtual columns #12241 @ wk989898

https://docs.pingcap.com/tidb/stable/release-8.5.4

### DDL puller close panic fixed

- Affects: TiCDC setups that close the DDL puller

Closing the DDL puller could cause a panic; after this release, closing it no longer triggers that panic.

> - Fix a panic that could occur when closing the DDL puller #12244 @ wk989898

https://docs.pingcap.com/tidb/stable/release-8.5.4

### Azure Blob Storage changefeed task stalls fixed

- Affects: TiCDC changefeeds using Azure Blob Storage as the downstream

Tasks could get stuck in this setup; this release addresses the issue.

> - Fix the issue that changefeed tasks might get stuck when using Azure Blob Storage as the downstream #12277 @ zurakutsia

https://docs.pingcap.com/tidb/stable/release-8.5.4

### TiCDC replication of `DROP FOREIGN KEY` DDL fixed

- Affects: TiCDC setups replicating DDL to a downstream (Inferred)

When `DROP FOREIGN KEY` DDL was replicated to a downstream, it was not sent. It is now replicated.

> - Fix the issue that DROP FOREIGN KEY DDL is not replicated to the downstream #12328 @ 3AceShowHand

https://docs.pingcap.com/tidb/stable/release-8.5.4

## Enhancements (11)

- gRPC overload wait-time metrics
- Per-table distribution inspection and redistribution (To use: v8.5.4 or later)
- TiDB adds global index support for non-unique columns in partitioned tables (To use: TiDB v8.5.4 or later)
- Optional redesigned TiCDC architecture (To use: Set the TiCDC configuration item `newarch` to `true`.)
- SELECT-list subquery decorrelation control variable added
- `EXISTS` subquery rewrite control variable added
- Embedded `ANALYZE` during `ADD INDEX`
- Support for `semi_join_rewrite` on Semi Joins in IN subqueries (To use: The `semi_join_rewrite` hint, applied to Semi Joins in IN subqueries)
- Backoff metric added to the Grafana SQL Execute Time Overview panel
- PD table-level Region scattering across scatter-role and engine dimensions
- TiCDC support for filtering unsupported DDL types (To use: The `ignore-txn-start-ts` parameter in the filter configuration)

## Dependency updates (1)

1 more, none mentioning security

- `- Upgrade the Golang version from 1.23.0 to 1.23.12 and update related dependencies #9788 @ JmPotato`

## Other changes (50)

- Repeated compaction tasks queued to `cleanup_worker`
- SST ingestion compatibility guard behavior
- Titan region-size estimation issue noted
- IMPORT INTO cleanup failure issue listed
- Slow `Tick(CompactCheck)` handling observed
- Slow reads with the resource-control auto priority limiter
- Async snapshot tail latency with growing SST-file counts
- BR restore download-failure logging issue
- Proposed increase to the default RocksDB manifest size
- Frequent store-size reporting issue recorded
- SST importer cache deadlock issue noted
- tikv-cdc assertion panic recorded
- `RawRecords` top-N aggregation proposal
- TiKV RocksDB write crash issue listed
- Graceful shutdown for TiFlash MPP tasks
- TiDB's default wait before querying new TiFlash nodes removed
- TiKV increases the default for `rocksdb.max-manifest-file-size`
- TiCDC adds the `scheduler.region-count-per-span` table-splitting setting
- TiDB accepts over-scale `DECIMAL` inserts with truncation
- TiDB estimation strategy optimized
- TiDB optimizer index-selection logic adjusted
- TiDB query estimation optimized for columns with few distinct values
- Optimized estimates for TiDB Index Join queries with `LIMIT OFFSET`
- TiDB out-of-range estimation optimized
- TiDB audit log now includes statement ID information
- Raft GC checks split into two phases
- Region MVCC GC handling moved to GC worker threads
- gRPC thread pool size now scales with CPU quota by default
- PD emits fewer unnecessary error logs
- TiFlash skips unnecessary reads during TableScan
- TiFlash TableScan performance optimized for wide, sparse tables
- Reduced TiFlash CPU overhead from adding vector indexes
- Reduced unnecessary TiFlash logs for useless Raft commands
- TiFlash `TableScan` performance improved on small partitioned tables
- Case-insensitive upstream `GTID_MODE` matching in TiDB Data Migration
- `USE INDEX` hint failure with TiFlash isolation reads fixed
- TiDB Decimal handling in prepared statements fixed
- Flow control fix for excessively large L0 SST ingestion
- Global sorting fixed for a 1:2 CPU-to-memory ratio
- Pending DXF task cancellation fixed
- Instance-level plan cache enabled for `select 1 from dual`
- Planner failures when query join order changes fixed
- `ONLY_FULL_GROUP_BY` negative-value validation failure fixed
- Incorrect plans from the DP join order algorithm fixed
- Errors from concurrent `ADD COLUMN` and `UPDATE` fixed
- Merge Join filter-condition cost calculation fixed
- PD Client retry strategy initialization fixed
- TSO Follower Proxy error handling fixed
- TiFlash query failures fixed
- Restore point creation failure with special schemas fixed

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
