The FCV upgrade/downgrade procedure can fail in ways that require a support case (e.g. abortAllReshardCollection() failing with ManualInterventionRequired). This means that the config server would be stuck in the transitional FCV while the shards are still in the original FCV. Since some feature flags are enabled in the transitional FCV, the cluster could be behaving differently between the config/shards. Therefore, we should be testing that features can work if setFCV fails in the middle of transitioning.

The goal of this ticket is to provide additional coverage for testing robustness of the setFCV command (for example, when it randomly fails).