The Pipe and Filter Architecture, while having specific use cases where it excels, such as batch processing and data transformation pipelines, is often not the preferred choice due to several factors. One of the main reasons is its limited flexibility in handling complex interactions and dynamic behavior between components, which are common in modern applications that demand high interactivity and real-time processing.

Moreover, in this architecture, each filter (or processing unit) is typically stateless, which can make it challenging to manage stateful operations or to coordinate tasks that require shared state or context, decreasing its appeal in systems that need such capabilities. Contemporary application requirements often involve intricate dependencies between system components that don’t map well onto the linear flow of pipes and filters.

Additionally, performance can be a concern in scenarios where data volumes are exceptionally high or where low-latency processing is critical. The architectural overhead of passing data through multiple filters can lead to inefficiencies compared to more centralized or integrated approaches.

The evolving trends in software architecture also play a role. Architectural styles such as microservices, event-driven architectures, and reactive systems provide more granular control, scalability, and resilience features that align better with current cloud-native and distributed processing requirements. These paradigms support sophisticated orchestration and real-time data processing, which is often beyond what Pipe and Filter can efficiently handle.

In summary, while Pipe and Filter Architecture offers simplicity and a clear separation of concerns, its lack of adaptability to modern application needs, difficulties handling stateful operations, potential performance bottlenecks, and a general shift towards more flexible and powerful architectural patterns contribute to its declining popularity.