Software development pipeline tools construct pipelines by combining tools, files, and other resources, to build, integrate, test, deploy, or otherwise implement operational functionality in computing systems. Some pipelines are simple, but others are stochastic due to conditional execution, task addition or removal, resource availability randomness, and other causes. Some stochastic pipelines also include a hierarchy with multiple levels of task groupings, which adds complexity. Pipeline performance optimization uses critical paths, but critical paths are challenging to identify in stochastic pipelines. Tools and techniques are presented to automatically identify likely or actual critical paths and to indicate constituent critical tasks as improvement options for stochastic pipelines in software development or other industrial activities. Pipeline representations include directed acyclic graph data structures of constituent tasks. Computationally applying relevance filters helps identify performance improvement options based on historic execution data, without requiring the predefined task dependency information that stochasticity prevents.