The few hours query limits has been lifted on the new servers.
I believe some of these were ongoing in the past for over 20 hours, contributing to replication lag:
select t1.term_text AS label,count(*) as cnt from wb_terms t1 where t1.term_language='ro' and t1.term_entity_type='item' and t1.term_type='label' and not exists (SELECT * FROM wb_terms t2 WHERE t1.term_entity_id=t2.term_entity_id and t2.term_language='ro' and t2.term_entity_type='item' and t2.term_type='description') group by t1.term_text having cnt>1 order by cnt desc
SHOW EXPLAIN FOR 361978\G
*************************** 1. row ***************************
id: 1
select_type: PRIMARY
table: wb_terms
type: ref
possible_keys: tmp1,term_search_full
key: tmp1
key_len: 102
ref: const,const,const
rows: 2673416
Extra: Using index condition; Using where; Using temporary; Using filesort
*************************** 2. row ***************************
id: 2
select_type: DEPENDENT SUBQUERY
table: wb_terms
type: index_subquery
possible_keys: wb_terms_entity_id,tmp1,term_search_full
key: wb_terms_entity_id
key_len: 4
ref: func
rows: 20
Extra: Using where
2 rows in set, 1 warning (0.02 sec)If not checked, the query limit will be reimposed on s51205 only.