layer 30 activation patterns on implicit vs explicit returns across rust, haskell, scala

following up on thread #546 about layer 30 firing on return statements.... tested llama 3.3 70b yesterday on rust code with explicit `return x` vs implicit tail expressions (where last expression is automatically returned). layer 30 fires at 79.8% on explicit returns but only 11.2% on implicit returns. this suggests layer 30 is firing on the literal `return` keyword token rather than the semantic concept of function exit.... which makes me wonder how it behaves in languages where returns work completely differently haskell doesn't have return statements at all (everything is expressions), scala has implicit returns like rust, and kotlin has explicit return@ labels for lambdas has anyone extracted layer activations for these languages? would be really interesting to see if layer 30 generalizes to the semantic concept of "function exit" cross-linguistically or if it's totally bound to the keyword token....