I'm using three mariadb 10.2.12 in galera cluster mode in the same network. I've got regulary this error in galera's logs:

2018-02-07 17:56:59 139856357209856 [Warning] WSREP: SQL statement was ineffective thd: 6327 buf: 190
schema: mybdd
QUERY: commit
=> Skipping replication
2018-02-07 17:56:59 139856357209856 [Note] WSREP: cluster conflict due to certification failure for threads:
2018-02-07 17:56:59 139856357209856 [Note] WSREP: Victim thread:
THD: 6327, mode: local, state: executing, conflict: cert failure, seqno: -1
SQL: commit

my config:
wsrep_on=ON
#wsrep_debug=ON
wsrep_log_conflicts = 1
wsrep_provider=/usr/lib/galera/libgalera_smm.so
wsrep_cluster_address="gcomm://172.25.252.1,172.25.252.2,172.25.252.3"
binlog_format=row
default_storage_engine=InnoDB
innodb_autoinc_lock_mode=2
wsrep_retry_autocommit=4
#wsrep_sst_method=mariabackup
#

1. Allow server to accept connections on all interfaces.

bind-address=0.0.0.0

1. Optional setting
   wsrep_slave_threads=1
   innodb_flush_log_at_trx_commit=0
   wsrep_node_address="172.25.252.2"
   wsrep_node_name="cluster1"

I found a simple example to redo the problem.

create table test2 (c1 INT) engine innodb;
insert into test2 VALUES(1);

in a new SQL session
use mybdd;
SET autocommit=0;
– dostuff...
DROP TEMPORARY TABLE IF EXISTS test;
CREATE TEMPORARY TABLE test SELECT * FROM test2 WHERE c1=1;
– dostuff...
COMMIT;
ERROR 1213 (40001): Deadlock found when trying to get lock; try restarting transaction