This seems extreme especially if they're going for bare-metal. Even a single one of these DB servers would handle a lot of traffic if they're bare-metal.
When it comes to running Postgres on 70-100 servers I'm also not sure, unless they're doing some sharding at the application level, I'd expect the overheads of replication and resulting network traffic to be insane if they're merely replicating across all of them.
Their whole website should be able to run on a handful of these machines; their main cost and resource usage would be hosting & converting uploaded media, not the DB of app servers.
Maybe they are trying to bring up multiple sites (ie. AWS regions) for redundancy. If you have 10 "regions" then maybe hardware requirements look a bit more reasonable (7-10 DB servers each, etc.)
When it comes to running Postgres on 70-100 servers I'm also not sure, unless they're doing some sharding at the application level, I'd expect the overheads of replication and resulting network traffic to be insane if they're merely replicating across all of them.
Their whole website should be able to run on a handful of these machines; their main cost and resource usage would be hosting & converting uploaded media, not the DB of app servers.