When Should You Adjust Standard Errors for Clustering, Alberto Abadie, Susan Athey, Guido Imbens, and Jeff Wooldridge, in the Quarterly Journal of Economics, 2023, right here: https://academic.oup.com/qje/article/138/1/1/6750017.
This interesting article raises the issue of experimental content in a fresh way. Among its contributions, it distinguishes between standard errors under the assumption that you have observed all the relevant clusters in the population vs. a random sample of clusters from a much larger population. The standard clustering formula assumes the latter, but researchers are often in the former position, for example when they observe data from all states in the U.S., which is taken to be the relevant population. Using the standard formula in this situation can give highly inflated standard errors.
I was always suspicious of this standard approach to clustering (though I never was forced to stake out a clear position, and never did). This paper articulates one of the reasons why.