> It's fractionally slower, true, because of the serialization hit (string to binary). And with JSON columns you have to serialize on accesses, which is a lot

> It's fractionally slower, true, because of the serialization hit (string to binary). And with JSON columns you have to serialize on accesses, which is a lot slower in the read-mostly tests.