Galera Cluster is a great and powerful solution. However, issues could happen and we're going to see here how to resolve them.
First of all, and as always, look at the logs (syslog by default); important information is written there and most of the time the problem will be explicit!
Sometimes, during a very high load write demand, the Cluster may stop replicating during this phase. Generally, this is not a good situation and you should avoid this kind of query as much as possible.
One solution can be to use an intermediate software layer to avoid it, for example, a Advanced Message Queuing Protocol (AMQP) software such as RabbitMQ or ZeroMQ. The setup of this kind of tool is out of the scope of this book.
To detect the paused time replication, you need to look at the flow control status. If it's equal to 1, that means the replication is paused; if it's 0, then it's ok:
MariaDB [(none)]> SHOW STATUS LIKE 'wsrep_%'; +----------------------------+---------...