How probabilistic data structures can help us mitigate a common anti-pattern often encountered in stream processing applications: carrying non-aggregative raw data deep down into a streaming topology for calculations, such as a distinct count of elements.