The balancer_should_return_random_migrations_failpoint.js test exercises that we do random migrations if the balancerShouldReturnRandomMigrations fail point is enabled. To make those migrations fast it sets reshardingMinimumOperationDurationMillis and balancerMigrationsThrottlingMs to zero as config server parameters.

However the test also sets replica set parameters which don't include those two resharding/balancer parameters. Under the config shard topology (sharding_config_shard suite) the replica set parameters override the config server parameters, so this test runs slower than expected and can time out.

Suggested fix: Set the same options i.e. (reshardingMinimumOperationDurationMillis and balancerMigrationsThrottlingMs to zero) as both replica sets and config server parameters.