Oh man.. The cost to run a node now is almost 10x less than it was when I started witnessing on steem... Fuckin' spent a grand a month on servers almost at one point.. That was not fun at all.
You can run a basic node to carry out transaction broadcasts and to scrape blocks off of on a system with a bit less than 8GB RAM now adays, however the time it takes to get these economy setups going is measured in days due to the nature of the data replays.
Not sure how many transactions ETH processes but BTC is no where near billions of daily transfer operations, it's not even capable of such operation bandwidth as far as I know unless bitcoin core has had serious improvements since I last dug into it.
Hive is a distributed ledger.. As for if one can call a network where the vast majority of transactions are processed by the (usually) same 20 servers decentralized is something I'll leave open to everyone to think about (personally I think it's not nearly as optimal as it could be). Over half of all BTC mining occurs in China which in theory could lead to a chinese 51% attack if they all worked together.. While BTC and ETH are better decentraliztion wise neither of them is so decentralized they are untouchable.. :/
As far as API calls being indexed by external databases that is one of the few ways to make calls to fetch old data seamless from a users perspective. It takes less time to query a localized DB than it does to have the hive API go and search for it.. From a technical standpoint it makes sense.. Think of HIVE as a hard drive and the hivemind as a sort of RAMdisk.
Will have a look at these projects you mention. Always interested to see what type of new and innovative shit guys are coming up with!
RE: For Hive to succeed, stakeholders need to not be completely delusional cultists