Transactions that touch a single shard that contains an arbiter or contains a node with enableMajorityReadConcern:false will be allowed in 4.2. We should test that:
- A single shard transaction can succeed against a replica set that contains an arbiter.
- A single shard transaction can succeed against a replica set whose primary has enableMajorityReadConcern:false and whose secondary is lagged. We should test this to make sure that the transaction is selecting a read timestamp based on lastApplied, as opposed to starting at the majority commit point. If it started at the majority commit point, we would not be able to obtain a snapshot, since we won't keep history back that far.