The existing causally consistent transaction passthroughs are mostly meant to ensure enabling causal consistency doesn't break the core transaction functionality, so they run against 1 node replica sets (e.g. sharded_causally_consistent_jscore_txns_passthrough). They also do so because currently transactions cannot read from secondaries. To better test causal consistency before transactions can run on secondaries, these suites could be updated to use multiple node replica sets and run all non-write commands that are not part of a transaction with secondary read preference and afterClusterTime read concern.