Does Cassandra offer any semantics similar to REPEATABLE READ?

I believe you can do this within a “batched” LWT, but doesn’t this incur significant network overhead due to PAXOS? REPEATABLE READ allows other concurrent transactions to make some progress and not be completely blocked. Doesn’t batched LWT completely block all other concurrent batch LWTs ?

We take REPEATABLE READ for granted on RDBMS, yet seems to be really costly on Cassandra. What techniques do developers use at the application level to enforce REPEATABLE READ on top of Cassandra ? (mainly to avoid Phantom Reads and Lost Updates)