We're seeing a few errors due to large transactions on S5. Unsure yet if this is Wikidata in particular or a general MW core issue, however all the recent examples are from wikidatawiki.
Wed Feb 18 6:05:39 UTC 2015 mw1011 wikidatawiki RecentChangesUpdateJob::purgeExpiredRows 10.64.32.28 1197 \ Multi-statement transaction required more than 'max_binlog_cache_size' bytes of storage
max_binlog_cache_size = 4GB. It's technically possible to increase this limit, but it would be unwise to do so since large transactions are going to cause replication lag. Whatever job generates this query really needs improved batching with more frequent commits.