Also he could consider adding in some fancy redundancy (SD-WAN, BGP, IP-SLA, whatever) to ensure that a single point of network I/O failure does not cascade to

Also he could consider adding in some fancy redundancy (SD-WAN, BGP, IP-SLA, whatever) to ensure that a single point of network I/O failure does not cascade to a split-brain scenario. Might reduce bandwidth or increase latency, but a total outage is pretty painful when you're taking about distributed/replicated databases.