I'd like to see their migration strategy as well. I mean, they say moving from pgsql 9.2 to higher version (which then allows online upgrades) is too much work. Yet they'll have to migrate to mysql, which will take much more engineering effort. For anything close to realtime, they'll need to copy the old data, while at the same time forking the new writes into both pgsql slaves and new mysql servers. And they cannot use WAL for that without some advanced processing.

I hope this follows in the next blog post.