I don't think those physicists are making any Gaussian assumptions, right? They're computing a p-value as normal (how often would my test statistic generated under the null be more extreme than observed), and then using the Gaussian distribution as a scale to communicate how small their p-value is.