Interesting analysis! Do you also have insights into which JSON Schema constructs are used most often in that dataset?
There have been research in that area (i.e. https://link.springer.com/chapter/10.1007/978-3-030-34146-6_...) but usually the input set is too small or biased.
As a JSON Schema TSC member, I consulted with various organisations on large scale JSON Schema use, which is where things get truly interesting (>8000 complex interconnected schema data models) though all of them are sadly closed-source.