As a fairly staunch advocate of PostgreSQL for the 99% use case, I'd be really interested to hear what makes RethinkDB stand out?

The admin UX... setting up a redundant sharded cluster with auto-failover in RethinkDB is in the box, straight forward and easy. With PostgreSQL you're stuck choosing between at least a half dozen less than optimal solutions.

That would be the single biggest difference... Neither have particularly great DBAAS offerings, outside of RDS Postgres, though if you're on Amazon, you can use Compose.io's rethinkdb option there too.

RethinkDB has sharding + redundancy similar to ElasticSearch or Cassandra, while having some other features that put it more in line with an SQL rdbms, while having an administrative UI that's second to none.. and I mean none.

Combine the best ideas from a relational database, a document store, sprinkle in some realtime change feeds, and an awesome admin interface. About all I need for a modern web application.

RethinkDB is also my default at this point, their scaling features are great. But they have several major unresolved issues, check their issue tracker on github for reference.

Anything serious that can corrupt or eat data?

No not really, It's really solid. But stuff like correlating writes with feeds are not possible atm.

One thing we've done in the past is included the feed name / source in the document (just an fyi...not sure if applicable to your use case or not).