Quorum in XtraDB Cluster - Monitor from Split-Brain

Hi support,

I have 3 PXC nodes running on CentOS x64 servers and 2 GARBD (located in 2 HAProxy Load Balancers in front, CentOS x64 too), so in total I have 5 odd numbers of PXC nodes.
If one of these nodes (1 of 2 garbd or 1 of 3 PXC) will fail (Ethernet link down, power failure, system OS errors, etc) I will have 4 PXC nodes and so on if will fail 3 PXC nodes, in such case it will Split-Brain ?
How to avoid such situation?

If 2 PXC nodes will go down at the same time and I will have 2 GARBD and 1 PXC node?

If one of 5 nodes go down, you won’t have split brain because those 4 have quorum. 4 against 1. The same applies if two nodes go down. 3 vs 2, so those 3 still have quorum.