Αdded support for asynchronous execution of ForeignScan operations under MergeAppend, controlled by the enable_async_merge_append parameter, which is enabled by default. If the operations under MergeAppend support asynchronous execution, requests are sent asynchronously at the start of the MergeAppend operation, and the results are cached as they are received. These cached results are then used, just as they would be in synchronous MergeAppend, for merge sorting.

Implemented the ability to use sorting on the remote server if it allows performing MergeJoin or MergeAppend operations. This is controlled by the postgres_fdw.additional_ordered_paths parameter, which is enabled by default in new installations but must be explicitly enabled in upgraded clusters.

Added support for the limit clause pushdown under Appendand MergeAppend when there is a Sort plan node between LIMIT and Append. It is possible when rows in subplans of Append/MergeAppend are already sorted in the necessary order.

Sped up INSERT, UPDATE, and DELETE operations with global tables. Added the shardman.gt_batch_size configuration parameter that specifies the buffer size for INSERT and DELETE commands executed on global tables.

Added a limitation on creating sharded and local partitioned tables based on the same attribute.

Added a new shardman.broadcast_query function that returns an executed SQL statement results.

Added a new field CSNXidMap_last_trim to the shardman.pg_stat_csn view that shows the last time when the shardman.trim_csnxid_map() function was called.

Improved the state consistency checks for the shardman application.

Fixed an issue with inappropriate resource allocation, which could cause errors in some corner cases when tuples were spilled to disk.

Fixed a bug in pg_rewind that previously resulted in the former primary server full resync on replica promotion.

Upgraded supported version of pgpro_pwr to 4.7.

Added logging of the updated parameters in case it results in postgresql restart.

Improved the logic for obtaining the state of the PostgreSQL instance.

Improved shardmand log messaging.

Fixed a bug that previously resulted in the shardmanctl psql command failure.

Added support for compression level values depending on the compression algorithm when creating a backup with shardmanctl probackup backup.

Updated the shardmanctl benchmark with a new dependency between the pgbench_branches number of records and the the number of nodes. This allows a better distribution of data between nodes.

Added the shardmanctl shard reset command that resets nodes of a replication group if they are in a state of hanging.

Added the shardmanctl daemon set command that allows updating the log parameters without restart.