Performance of the Web Audio graph is heavily dependent on skipping processing of silence because processing in unreferenced nodes continues until they are garbage collected. Nodes should produce a null chunk when they know the output is silent. If not, the rest of the graph downstream will need to process the chunk.