Upon investigation, decided this to be a compiler bug
(happens with new compiler, on code that did not change for the last 15 years)

Fixed by de-optimizing single function remove_key(), using MSVC pragma

In dict_index_t::clear(), InnoDB frees all the page except root page.
root page leaf segment has reset and does reinitialize again.
t in fseg_create(), we do have the assumption that only
FIL_PAGE_TYPE_TRX_SYS or FIL_PAGE_TYPE_TRX_SYS page should
be re-created for non-full-crc32 format. This assumption is wrong
in case of rollback of bulk insert operation.

MDEV-27006 Assertion `!lock_trx_has_sys_table_locks(trx)' failed in dberr_t row_discard_tablespace_for_mysql(dict_table_t*, trx_t*)

- InnoDB import operation fails when it tries to unlock data
dictionary lock before releasing the lock on system tables.

The debug assertion that was added in
commit 9b967c4c
tripped Valgrind and MemorySanitizer.

buf_block_init(): Assert that block->page.hash was zero-initialized.

This fixes up commit d22c8cae

The previous threads locked need to be released too.

This occurs if the initialization of any of the non-first
mutex/conditition variables errors occurs.

Fix a bug of unreleased lock ctrl_mutex in the method create_worker_threads

added summary/description per package.

MDEV-26956 LeakSanitizer/Valgrind errors in trx_mod_table_time_t::start_bulk_insert upon adding system versioning

 InnoDB fails to apply buffered insert operation for
'mysql/transaction_registry' table during system versioning DDL.
To avoid this, DDL calls extra(HA_EXTRA_IGNORE_INSERT) to
inform the InnoDB for applying the buffered insert operation.