As part of the BF fix made in SERVER-133105 team noticed that in FSM tests execution, when the tests are executed in add remove shards suite if any failures in the shard transitioning appear, the setup function is called repeatedly without cleaning up the progress made on previous call, which can lead to duplicate entries appearing if the setup function in fsm is not idempotent

Ideally the fix to this problem should be as part of the retry function which is triggered on errors encounter to run teardown() before re-invoking setup(), but that would be a big change as it would affect every single existing fsm test. Additionally it's known that some of them don't have teardown() function specified at all.

As part of this ticket we should investigate how big is the problem (how often does it occur) with setup function reentrance when fsm test is run in add remove shards suite and what might be our best solutions to this problem (how wide the change would be if we design the fix around invocation of teardown() function/how many fsm tests would need to be rewritten etc) and try to tackle the problem.