Guess what? The most used lemma in the Stacks project is currently Nakayama’s lemma with 73 references. Then there is a gap for a while. The next most highly cited lemmas are:

- flat finitely presented morphisms are open with 38 references,
- a characterization of finite type morphisms with 33 references,
- descending properties of modules from a standard open covering with 32 references,
- pushforward of quasi-coherent modules along quasi-compact and quasi-separated morphisms are quasi-coherent with 26 references,
- cohomology of quasi-coherent modules on affines is zero with 25 references,
- a quasi-compact algebraic space has an etale covering by an affine scheme with 25 references,
- a scheme of finite type over a Noetherian scheme is Noetherian with 24 references,
- and so on.

I guess it makes sense that the first few of these would be results that we consider to be part of the language, and somehow not really results at all (if that makes sense).