Partial repaint is too effective, and we'd like to be able to measure performance without carefully structuring the benchmarks. For example, right now partial repaint is culling any blurs in the multibackdrop case, which we should be using to track  https://github.com/flutter/flutter/issues/132735