I would certainly consider Roslyn analyzers capable of covering "a tiny fraction" of possible errors :) They are quite capable of covering for many common types

I would certainly consider Roslyn analyzers capable of covering "a tiny fraction" of possible errors :) They are quite capable of covering for many common types of structural coding mistakes.