The manner in which stuck lambdas are identified is unclear. The preference (and current implementation) is do so on a per-argument basis, even if the expression is many levels deep. Thus, a stuck lambda or method reference nested inside of a conditional or another lambda (as the lambda's return expression) "freezes" the whole argument expression -- the granularity of stuckness is per-argument, not per-subexpression. In contrast, a stuck lambda nested inside another generic method call does not "freeze" the whole method call.