BF-7458 manifests as a spontaneous reelection of primary during a database operation between tests such as checkUUIDsConsistentAcrossCluster (which cache primary host name).
This reelection is not expected by the tests and hence causes them to fail.
There are 2 possible solutions assuming that reelections are normal and can happen due to heartbeat delay.
1. look into all failing cases and use the current primary when making requests.
2. configure the sharded cluster in a way that reelection will choose the same primary.