Is there an updated list similar to the one here:
That contains all the possible variables and their possible settings (local/global etc)? I am especially interested to learn which of the diagnostic options are or should be dynamic as having problems with a 3 node XtraDB cluster remaining up and it is failing node after node in rapid succession with little or no diagnostic output. It is completely standard.
The transaction are individually not huge.
The number of transactions is c250 queries per second.
The logs are saying various things but mysqld (lose the cluster node completely if that was the cause), mysql or bash are invoking the oom-killer depending what is running