Percona / Galera - 110 WSREP error - which way to dig?

Question

linux

Cluster for >10 servers evenly distributed between DCs, loaded system.
The logs often show: WSREP: Failed to report last committed XXXX, -110 (Connection timed out)
Omitting periodic SST / IST (managed, during maintenance) - in which direction to dig to come to improved stability?
The cluster is working, the systems are working - that is - these are just warnings - it has little effect on stability.
Could this be the cause of bunches of deadlocks in the database, or are there subtle problems on specific nodes?
In addition, the logs are generally clean.