This guy doesn't really get it. I assume he's talking about "high transaction volume" when he's talking about scalability. Anyone who knows anything about scaling to high transaction volumes knows that it's not SQL that gets in the way, it's ACID. Transactions don't scale regardless of whether they're implemented in a SQL database or a NoSQL database. This guy's VoltDB project is still shackled by ACID properties, and I don't really see the advantage of it over using MySQL with the in-memory storage engine.
tl;dr -- a lot of NoSQL ("NewSQL!?") information is FUD.
There may be FUD here, but I don't know if you understood Stonebraker's argument.
The whole point of VoltDB is that it is (supposed to be) dramatically faster than in-memory MySQL. That's not too hard to imagine, considering that MySQL uses one machine and VoltDB uses a cluster.
In fact, the whole point of NewSQL in general is that it is possible to implement SQL and ACID on a cluster and be dramatically faster than MySQL. While NoSQL is high-performance non-ACID non-SQL on a cluster, NewSQL is high-performance ACID SQL on a cluster; they're not the same thing.
tl;dr -- a lot of NoSQL ("NewSQL!?") information is FUD.