The population must be normal for any test to work.

Sample means will be approximately normal even if the population is not, provided $n$ is large.

Large samples are less accurate than small samples.

The standard deviation of the sample equals the standard deviation of the population.