A/A tests (also known as Null tests) are useful to validate that users are assigned to the control and experiment groups without bias.
Offline resampling methods, like bootstrapping, are better if you're looking to robustly estimate the variance of the experiment.