Insert and update delays in Galera Cluster

Hello

OS: Cent 7
DB: Maria 10.6.15 (Galera Cluster)
We are operating the service with the above specifications Galera Cluster (3 units).

Time of problem occurrence 00 ~ 02

■ Phenomenon
Around 00:00, a single Insert query and Delete were performed and stopped without being performed. After that, all Insert queries were delayed.

(The tables of the insert and delete statements are different tables.)

From 00:34, locks started to occur when updates were received using the same pk value as a parameter.

At 00:41, DB 01 of Galera Cluster was removed.
The following contents were recorded in the log at that time.

[ERROR][FATAL] InnoDB: innodb_fatal_semaphore_wait_threshold for dict_sys.latch was exceeded. See https://mariadb.com/kb/en/how-to-produce-a-full-stack-trace-for-mysqld/ .

I restarted the problematic database server and reconnected to the Galera cluster.

If anyone has experienced a similar situation or knows how to fix it, please share.