fastFail will temporarily remove a cassandra host from the connection pool if it experiences "enough" errors: the failure detection algorithm used (by Finagle) for fastFail is similar to the one used internally in cassandra for gossip.
If you'd like to survive errors on individual hosts, you'll need to enable retries, since they are disabled by default:
...retries(3).fastFail(true)...