Why is the chain's liveness and resiliency to server failures dependent on where people happen to be keeping their validators? This isn't a problem in Ethereum 1.0.
I mean, if all the miners were on one datacenter and that datacenter went down, so would Ethereum 1.0. That's the same sort of failure you're worrying about for 2.0.
A blockchain's resiliency in part depends on how many full replicas are available -- if at least one replica exists somewhere, the chain can be resuscitated even if all block producers die.
In Ethereum 2.0, nodes don't maintain a full copy of the chainstate -- not even block producers do. Therefore, Ethereum 2.0's resiliency depends on the continued availability of N > 1 partial copies of the chainstate, such that the full chainstate can be reassembled from them. This is strictly less resilient.
