If I understood your blog post correctly (http://sobereva.com/272), (meta-)GGA functionals are generally recommended when there is a risk of static correlation in ground state calculation, while not really recommended for anything else. (Please correct me if I'm wrong)

But I also know that nowadays, range-separated functionals such as wB97M-V are very sophisticated. And if I think about the concept of range-separation, I don't really see how a good-enough range-separated functional should be any worse than (meta-)GGA functionals even for moderate static correlation. And if static correlation was really severe, DFT will not be suitable anyways.

I understand the merits of B97-3c and r2SCAN-3c, but my question is: Does a traditional (meta-)GGA functional like PBE or TPSS have any reason to be utilized for non-periodic, non-TM-cluster calculations if I only care about accuracy?