The following tests fail when shard servers are started as replica sets, each for a different reason.

migrateBig_balancer.js
Fails with the error

InternalError: allocation size overflow

startup_with_all_configs_down.js
This test restarts a shard without restarting mongobridge, but when the shard is a replica set the test fails with the error

Could not find host matching read preference

The following tests need more investigation as to why they fail:
bouncing_count.js
kill_pinned_cursor.js
move_stale_mongos.js
sharded_agg_cleanup_on_error.js