we can have a larger number of nodes, even if any one of them is relatively less reliable, where each node can take over and cover for any downtime of another node.
So we actually have this, there are something like 400+ active witness nodes, the list pages on most of the interfaces just don't show that many. Because below a certain number, they have almost no vote support nor do they cycle into rotation unless hundreds above them fail over, but they ARE there, paying for servers with little to no remuneration for their support of the chain's block processing power, speed and infrastructure.
But the pay scale is so tilted toward the top 20, that it causes a lot of churn in these lower level witness nodes, they come and go and due to lack of expiring votes, remain as disabled, and having to be accounted for on the list pages and the chain scheduling algo itself, discarding those who are not active anymore. There is no way for a witness to completely resign on their own, only for their voters to completely remove all weight from them. My own dead st33m witness hasnt run in years, but has voters,who likewise are unreachable (ive checked) who left votes on them and then disappeared completely, without removing those votes. That dead witness continues to clutter up the witness lists (way down in the 400s but still) and has to be accomodated as inactive in the chain scheduling code. Extra cycles that could be avoided by allowing a witness to cancel all votes on itself and "quit" if they need to or want to for any reason at all.
But yeah, in summary, we have hundreds of nodes - resilient and failover by design, but nobody knows they exist.