I'd posit its more a matter of maturing a new paradigm. There's a lot more edge cases you have to cover as NoSQL became more popular for production-at-scale.
SQL has decades of production maturation, and has wider domain knowledge.
I'm sure there's some of that. But a lot of the early problems were a bit more weighted towards poor engineering in general, IIRC. For example, I seem to recall an early problem was truncating large amounts of data on crash occasionally.
Not that that's really a surprise or was unknown, it's just fairly new to see in the open source ecosystem instead of the enterprise one.