replmonitor_bad_seed.js waits for the replica set nodes to be reachable but if a primary is not available (due to step downs for example), the write to the shard may fail:

    // Verify that the replSetMonitor can reach the restarted set
    awaitRSClientHosts(st.s0, replTest.nodes, {ok: true});

    assert.writeOK(st.s0.getDB('test').user.insert({x: 1}));

We should call ReplSetTest.awaitNodesAgreeOnPrimary on the replica set before executing the sharded write.