That's incredibly well written, constructive feedback by the Uber team! I wonder if the Postgres team has taken this on board and attempted to resolve some of the design issues, e.g.: by moving the cache in-process.
The Uber team betrays their lack of knowledge about PostgreSQL in the article, so take it with a huge grain of salt. I recommend reading Markus Winand's article.
But that does not mean though that there are not also real issues with table bloat in PostgreSQL, and they are being worked on by creating a new storage engine with a different MVCC implementation.
