The first thought that came to mind was "Wow !!! You must have rather big transactions."
According to mysqlperformanceblog.com's Reasons for run-away main Innodb Tablespace, these are the main issues that cause ibdata1 to grow:
- Lots of Transactional Changes
- Very Long Transactions
- Lagging Purge Thread
The undo logs inside ibdata1 will hold lots of info to support transaction isolation levels. Since you are using READ-COMMITTED, this growth can be more significant.
I have discussed this before in the DBA StackExchange
Aug 21, 2015: Transactional DDL workflow for MySQLJun 16, 2014: MySQL Index creation failing on table is fullApr 23, 2013: How can Innodb ibdata1 file grows by 5X even with innodb_file_per_table set?- ... and more
SUGGESTION #1
Since you are using MariaDB, please look into configuring the undo logs outside of the system tablespace. Please read the MariaDB docs on innodb_undo_directory and innodb_undo_tablespaces and go from there.
SUGGESTION #2
You may have to mysqldump are your data and load it into a new installation of MariaDB. See my posts in How do I shrink the innodb file ibdata1 without dumping all databases? for other ideas. Either way, this will take some work !!!